Regional variations in occupancy frequency distribution patterns between odonate assemblages in Fennoscandia

This is an electronic reprint of the original article. This reprint may differ from the original in pagination and typographic detail. Author(s): Title: Year: Version: Please cite the original version: All material supplied via JYX is protected by copyright and other intellectual property rights, and duplication or sale of all or part of any of the repository collections is not permitted, except that material may be duplicated by you for your research use or educational purposes in electronic or print form. You must obtain permission for any other use. Electronic or print copies may not be offered, whether for sale or otherwise to anyone who is not an authorised user. Regional variations in occupancy frequency distribution patterns between odonate assemblages in Fennoscandia Korkeamäki, Esa; Elo, Merja; Sahlén, Göran; Salmela, Jukka; Suhonen, Jukka Korkeamäki, E., Elo, M., Sahlén, G., Salmela, J., & Suhonen, J. (2018). Regional variations in occupancy frequency distribution patterns between odonate assemblages in Fennoscandia. Ecosphere, 9(4), Article e02192. https://doi.org/10.1002/ecs2.2192 2018 Regional variations in occupancy frequency distribution patterns between odonate assemblages in Fennoscandia ESA KORKEAM€ AKI, 1 MERJA ELO, 2 G€ ORAN SAHL EN, 3 JUKKA SALMELA, 4 AND JUKKA SUHONEN 5,  1 Water and Environment Association of the River Kymi, Tapiontie 2 C, FI-45160 Kouvola, Finland 2 Department of Biological and Environmental Sciences, University of Jyv€ askyl€ a, P.O. Box 35, FI-40014 Jyv€ askyl€ a, Finland 3 Ecology and Environmental Science, RLAS, Halmstad University, P.O. Box 823, 30118 Halmstad, Sweden 4 Regional Museum of Lapland, Pohjoisranta 4, FI-96200 Rovaniemi, Finland 5 Section of Ecology, Department of Biology, University of Turku, FI-20014 Turku, Finland Citation: Korkeam€ aki, E., M. Elo, G. Sahl en, J. Salmela, and J. Suhonen. 2018. Regional variations in occupancy frequency distribution patterns between odonate assemblages in Fennoscandia. Ecosphere 9(4):e02192. 10.1002/ecs2.2192 Abstract. Odonate (damselfly and dragonfly) species richness and species occupancy frequency distributions (SOFDs) were analyzed in relation to geographical location in standing waters (lakes and ponds) in Fennoscandia, from southern Sweden to central Finland. In total, 46 dragonfly and damselflyspecieswere recorded from 292 waterbodies. Species richness decreased to the north and increased with waterbody area in central Finland, but not in southern Finland or in Sweden. Species occupancy ranged from 1 up to 209 lakes and ponds. Over 50% of the species occurred in <10% of the waterbodies, although this proportion decreased to the north. In the southern lakes and ponds, none of the species occurred in all lakes, whereas in the north, many species were present in all of the studied waterbodies. The dispersal ability of the species did not explain the observed species occupancy frequencies, but generalist species with a large geographical range occurred in a higher percentage of the waterbodies. At Fennoscandia scale, we found that the unimodal satellite pattern was predominant. However, at smaller scale, we found geographical variations in odonate species SOFD patterns. The most southern communities followed the unimodal satellite-dominant pattern, whereas in other regions, communities fitted best with the bimodal core–satellite patterns. It seems that the richer species pool in the southern locations, and the larger distribution range of the northern species, skewed the unimodal pattern into a bimodal satellite-dominant pattern. Key words: core–satellite species patterns; damselfly; dragonfly; freshwater lake; Odonata; pond; species richness. Received 21 November 2017; revised 16 February 2018; accepted 20 February 2018. Corresponding Editor: Robert R. Parmenter. Copyright: ©2018 The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. E-mail: juksuh@utu.fi INTRODUCTION The shape of the species occupancy frequency distribution (SOFD) is a widely studied area in community ecology (see reviews by McGeoch and Gaston 2002, Jenkins 2011). In natural communities, many species occur either on few sites (satellite species; often rare species) or at many sites (core species; often common species), forming a bimodal core–satellite pattern (Hanski 1982, 1999). Using the core–satellite species pattern will yield more information about community structure than mere species richness. In terrestrial habitats, SOFDs generally have a bimodal core–satellite pattern (Hanski 1982, 1998, 1999, McGeoch and Gaston 2002, Jenkins 2011). By contrast, in aquatic communities only weak support for a bimodal SOFD pattern has been found (Verberk et al. 2010, Heino 2015). It therefore remains unclear whether or not aquatic ❖www.esajournals.org 1April 2018 ❖Volume 9(4) ❖Article e02192 animal communities exhibit this general macroecological pattern. In many cases, SOFD patterns can largely be explained by sampling methods and efficiency. For example, the grain size as well as the extent and intensity of sampling can vary extensively, and a decrease in sample area or number of sites may change the observed SOFD patterns from unimodal to bimodal (McGeoch and Gaston 2002). However, SOFD patterns may also depend on abiotic and biotic factors: (1) habitat disturbance, (2) niche breadth, (3) dispersal ability, (4) sampling site position within geographical range, and (5) geographical range size distribution (McGeoch and Gaston 2002, Jenkins 2011, Jokim€ aki et al. 2016). First, in stable habitats with low levels of disturbance, communities should follow the bimodal SOFD pattern (Jenkins 2011). Second, generalist species will occur at most sites, whereas specialist species will occur at fewer sites. This accounts for the clear nested species subset pattern observed for dragonfly communities (Sahl en and Ekestubbe 2001, Koch et al. 2014). Moreover, generalist species with broad niches tend to have a wide geographical distribution, whereas specialist species are limited by their narrow niches. Thus, this nichebased hypothesis mainly predicts a right-skewed unimodal SOFD pattern (Brown 1984). Third, dispersal ability varies between species, and in aquatic insects, it depends on body size (Heino 2015). Large species are often good fliers that are capable of dispersing over long distances, whereas smaller species can be expected to have a more restricted dispersal ability (Conrad et al. 1999, McCauley 2006, Wikelski et al. 2006, McCauley et al. 2008, Hassall and Thompson 2012, Troast et al. 2016). Thus, the dispersalability hypothesis predicts a satellite-dominant unimodal SOFD pattern (Collins and Glenn 1997). Fourth, it is traditionally thought that species tend to have their highest density/abundance at the center of their geographical distribution (Brown 1984). In addition, occupancy and abundance of species are often positively correlated (Hanski 1982, Brown 1984, Collins and Glenn 1997). Accordingly, as the species pool is richer, the relative number of satellite species can be expected to be higher in the southern than in the northern regions (McGeoch and Gaston 2002). Finally, regions at lower latitudes should have fewer core species and a larger number of satellite species than the regions at higher latitudes (McGeoch and Gaston 2002) because species occurring at lower latitudes have smaller latitudinal ranges than species at higher latitudes (the so-called Rapoport’s latitude rule; Gaston et al. 1998, McGeoch and Gaston 2002). Understanding the variation in the shape of SOFD patterns in aquatic communities needs further investigation. By using a semiaquatic insect group as study organisms, we will establish whether the bimodal SOFD pattern, which is often found in terrestrial taxa, is general also for aquatic communities. Moreover, as conservation efforts in a changing climate should be based on correct assumptions on current and projected community structure, we anticipate that a SOFD pattern analysis could be used as one of the cornerstones to such work. The specific aim of this study was to determine whether odonate SOFD patterns vary between the lakes and ponds of four geographical regions in Finland and Sweden, and whether the differences in SOFD patterns can be related to the five aforementioned abiotic and biotic factors. First, we predict that species richness increases with lake and pond area, since a larger area will in general allow for more niches (Oertli et al. 2002). Second, we predict that species occupation frequency in the lakes will increase with the geographical range of the species, because species with a wide geographical range are often locally abundant and therefore occur in many patches (Brown 1984). Third, we expect that generalist species occur more frequently in the waterbodies than do specialist species. Fourth, we predict that large-bodied species occur in a larger number of waterbodies than small-bodied ones, due to differences in dispersal ability (McGeoch and Gaston 2002, Heino 2015). Finally, we predict that southern regions have fewer core species and a larger number of satellite species than regions at higher latitudes (McGeoch and Gaston 2002). MATERIALS AND METHODS Description of data, methods, and study areas We used data from 292 lakes and permanent ponds (waterbodies from now on) in Sweden and Finland (Fig. 1) along a 900-km latitudinal extent. The original data have been gathered ❖www.esajournals.org 2April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. between 1995 and 2016 (Fig. 1) and consist of published studies as well as one previously unpublished survey (Table 1). All data were based on intensive field work (see more details in the original studies from Sahl en and Ekestubbe 2001, Flenner and Sahl en 2008, Honkanen et al. 2011, Suutari et al. 2009, Wittwer et al. 2010, Korkeam€ aki 2013, Koch et al. 2014, Suhonen et al. 2014, Al Jawaheri and Sahl en 2017). In all studied waterbodies, we aimed at detecting the majority of the species present, missing as few rare ones as possible. There have been publications in recent years discussing the reliability of various methods for estimation of odonate species richness (Raebel et al. 2010, Bried et al. 2012a,b, Hardersen et al. 2017), and it has been shown that even small samples could produce a fairly reliable species list for any given site, given that the sampling is repeated (Bried et al. 2012b). Further, it is well known that the rarest species at any site will always have less chance of being detected (Mao and Colwell 2005). Although the total latitudinal difference is only 8°, there is a profound climate and vegetation gradient between southern Sweden and central Finland. While oceanic climate and temperate broadleaf and mixed forests appear in southern Sweden, central Finland has continental climate in the mid-Boreal vegetation zone. Yearly mean temperature for the southernmost localities is above 7.0°C (SMHI 2017), while it is around 3.8°C for the northernmost ones (Finnish Meteorological Institute 2017). Further, there is an ecotone between the southern and the northern areas which constitutes the northern limit of a number of thermophilous tree species (Heikkil€ a and Sepp€ a 2003). All lakes are located in areas with numerous lakes (around one lake per 9km 2 ; Henriksen et al. 1998) of which we investigated only a small fraction. Thus, we divided the waterbodies into four groups based on their geographical location: southern Sweden (55°– 58°N, n=94), central Sweden (58°–61°200N, n=91), southern Finland (60°–61°300N, n=58), and central Finland (61°300–63°300N, n=49). We measured the geographical range of each species as the number of occupied 50 950 km squares in the maps by Boudot and Kalkman (2015), which represent an up-to-date compilation of known records in Sweden and Finland (up to 2014). In total, 20 of the species occurring in the study have much larger geographic ranges in Europe, in the Palearctic, or even in the Holarctic area. The species at the edge of their distribution vary between our areas: 7 in southern Sweden, 17 in central Sweden 8 in southern Finland, and 15 in central Finland. It is expected that more species are at their range margin further to the north. However, one species (Coenagrion johansoni) is at its southern border in southern Sweden. Note that several species are at their northern borders both in Sweden and in Finland. We also divided species into groups by their breeding habitat and dispersal ability. In regard to their breeding habitat, we used data from extensive field work in Finland and Sweden (Valle 1952, Korkeam€ aki and Suhonen 2002, Fig. 1. Location of the 292 studied waterbodies in southern Sweden (filled dot), in central Sweden (open dot), in southern Finland (filled dot), and in central Finland (open dot). ❖www.esajournals.org 3April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. Suhonen et al. 2010, 2014, Sahl en, unpublished data) to classify the species into three groups: generalists (14 species) breeding in both standing and running water (at least in Finland or Sweden), specialists (28 species) breeding mainly in standing water, and tourist species (five species) mainly breeding in running water but sometimes encountered in standing water (Table 2). The Odonata includes species with high as well as low dispersal capacity. It has been shown that some Anisoptera (dragonflies) have the capacity to move long distances, even across oceans (Troast et al. 2016, Alvial et al. 2017), while other Anisoptera have difficulties to pass a narrow landscape barrier such as a motorway ( Sigutov a et al. 2017). The same applies for Zygoptera (damselflies): Many species are poor dispersers (Watts et al. 2007), while others are capable of making long-distance migration, but probably more aided by wind than species in the other suborder. Suhling et al. (2017) showed that in a desert environment, specimens of Pseudagrion glaucescens were found more than 270 km from the nearest suitable habitat. Although less than the distances found for anisopteran species, this indicates that also some Zygoptera have high dispersal possibilities. One example is Ischnura hastata on the Azores, 3300 km distant from the nearest Caribbean population (Lorenzo-Carballa et al. 2017). We used the body size (measured as the mean value of minimum and maximum hindwing length; Dijkstra and Lewington 2006) as a proxy for dispersal ability. Statistical methods We used 10% occupancy classes and number or percentage of odonate species in each class to represent the geographical variation in occupancy frequency distribution, as recommended by McGeoch and Gaston (2002). We used Pearson correlation to test the relationship between species richness and waterbody area. As our localities varied from small ponds to relatively large lakes, the waterbody area was log 10 -transformed before analyses. We used Spearman rank correlation to test the relationship between a species’geographical range and its wing length, and generalized linear models with type III errors (negative binomial distribution; log link) to test the relationship between a species’geographical range and the number of waterbodies occupied. We used the same method to test differences in occupancy frequency between the different breeding habitat types. In this model, the breeding habitat type was set as a factor, and the geographical range, as well as its wing length, was used as continuous covariates. We tested differences between breeding habitat types within a Table 1. Number of waterbodies (lakes and permanent ponds) studied in Finland and Sweden. Region Number of waterbodies Waterbody area (ha) Methods SourceMean SD Min Max Southern Sweden 94 12.57 32.85 0.02 247.50 L Al Jawaheri and Sahl en (2017) L Wittwer et al. (2010) L Koch et al. (2014) L, E, A G. Sahl en, previously unpublished data Central Sweden 91 3.22 5.86 0.02 41.25 L, E, A G. Sahl en, previously unpublished data L Sahl en and Ekestubbe (2001) L Flenner and Sahl en (2008) Southern Finland 58 6.87 19.96 0.02 140.90 L Suutari et al. (2009) E, A Korkeam€ aki (2013) Central Finland 49 10.44 25.86 0.04 147.60 E, A Suhonen et al. (2014) L, E Honkanen et al. (2011) Combined 292 8.06 23.46 0.02 247.50 L, E, A Notes: The mean, standard deviation (SD), minimum (min), and maximum (max) are given for each of the four geographical regions. Methods indicate how the odonate species were sampled: L is larvae, E is exuviae, and A is adults. ❖www.esajournals.org 4April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. Table 2. A list of species observed in 292 waterbodies in Finland and Sweden. Species Suborder Habitat Wing length Range SS (n)CS(n)SF(n)CF(n) Total (n) Aeshna caerulea Anisoptera Specialist 80 508 ... 1... ... 1 Aeshna crenata Anisoptera Specialist 106 63 ... ... 21 1 22 Aeshna cyanea Anisoptera Generalist 100 363 17 19 2 38 Aeshna grandis Anisoptera Generalist 102 700 59 58 51 39 207 Aeshna juncea Anisoptera Specialist 95 770 18 73 39 42 172 Aeshna mixta Anisoptera Specialist 85 205 6 5 ... ... 11 Aeshna serrata Anisoptera Specialist 99 115 ... 3... ... 3 Aeshna subarctica Anisoptera Specialist 90 355 10 27 30 28 95 Aeshna viridis Anisoptera Specialist 87 88 1 2 5 ... 8 Brachytron pratense Anisoptera Generalist 72 210 14 5 2 1 22 Calopteryx splendens Zygoptera Tourist 61 315 1 ... 4... 5 Calopteryx virgo Zygoptera Tourist 58 640 4 ... 6... 10 Coenagrion armatum Zygoptera Specialist 39 315 1 5 11 7 24 Coenagrion hastulatum Zygoptera Generalist 40 663 42 75 26 49 192 Coenagrion johanssoni Zygoptera Specialist 36 370 ... 20 23 36 79 Coenagrion lunulatum Zygoptera Specialist 40 248 5 1 ... ... 6 Coenagrion puella/pulchellum Zygoptera Generalist 41 405 46 45 19 11 121 Cordulia aenea Anisoptera Generalist 68 615 45 55 33 44 177 Enallagma cyathigerum Zygoptera Specialist 38 618 26 15 10 17 68 Epitheca bimaculata Anisoptera Specialist 85 150 2 ... 41 7 Erythromma najas Zygoptera Generalist 43 495 46 28 23 28 125 Gomphus vulgatissimus Anisoptera Tourist 64 225 3 ... 0... 3 Ischnura elegans Zygoptera Specialist 35 333 44 ... 1... 45 Lestes dryas Zygoptera Specialist 45 270 ... 1... 12 Lestes sponsa Zygoptera Specialist 42 600 41 37 36 32 146 Lestes virens Zygoptera Specialist 39 78 1 ... ... ... 1 Leucorrhinia albifrons Anisoptera Specialist 60 323 9 13 17 14 53 Leucorrhinia caudalis Anisoptera Specialist 64 238 2 4 10 22 38 Leucorrhinia dubia Anisoptera Specialist 53 610 7 47 35 35 124 Leucorrhinia pectoralis Anisoptera Specialist 66 263 4 13 2 0 19 Leucorrhinia rubicunda Anisoptera Specialist 66 645 3 40 3 30 76 Libellula depressa Anisoptera Specialist 76 323 ... 2... ... 2 Libellula qadrimaculata Anisoptera Generalist 75 630 54 69 33 38 194 Orthetrum cancellatum Anisoptera Specialist 77 305 9 1 ... ... 10 Orthetrum coerulescens Anisoptera Generalist 60 165 3 2 1 ... 6 Platycnemis pennipes Zygoptera Tourist 45 340 1 ... 1... 2 Pyrrhosoma nymphula Zygoptera Generalist 44 415 28 9 ... 441 Somatochlora arctica Anisoptera Generalist 68 385 ... 1... 23 Somatochlora flavomaculata Anisoptera Specialist 76 290 6 12 4 9 31 Somatochlora metallica Anisoptera Generalist 78 743 27 17 11 27 82 Sympecma fusca Anisoptera Specialist 40 115 1 1 ... 02 Sympetrum danae Anisoptera Generalist 46 603 14 36 34 21 105 Sympetrum flaveolum Anisoptera Specialist 55 405 3 1 10 13 27 Sympetrum sanguineum Anisoptera Generalist 55 245 24 12 ... ... 36 Sympetrum striolatum Anisoptera Specialist 58 190 1 3 ... ... 4 Sympetrum vulgatum Anisoptera Specialist 60 420 6 12 4 ... 22 Notes: For each of the species, the following information is presented: suborder [Zygoptera (damselflies), Anisoptera (dragonflies)], breeding habitat (generalist is breeding in both standing and running water, specialist is breeding mainly in standing water, tourist is mainly breeding in running water but rarely also in standing water), mean hind wing length (mm; from Dijkstra and Lewington 2006), geographical range (1000 km 2 , geographical range area in Finland and Sweden from Boudot and Kalkman 2015), and number of waterbodies (n) from which each species was collected in the four respective regions; southern Sweden (SS), central Sweden (CS), southern Finland (SF), central Finland (CF), and combined data (Total). Note that Coenagrion puella and C. pulchellum were pooled, since they are inseparable as larvae. ❖www.esajournals.org 5April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. given geographical range with Kruskal–Wallis test, as data were not normally distributed. We applied the multi-model inference approach to the regressions of empirically ranked species-occupancy curves (RSOCs; Jenkins 2011). For these analyses, each of the data sets (combined data and the four different areas; species listed in rows and waterbodies as columns), was processed separately (Jenkins 2011). All analyses described below are based on relative occupancy (presence/absence) data at individual waterbodies. First, we calculated the sum of areas of all waterbodies where a given species was observed. We then divided each occupancy area by the total area of all waterbodies, to get the relative proportion of the total waterbody area that each species occupied (Hanski 1999). Second, we sorted the species by their relative occupancy values in decreasing order, where R i is the rank value for species i. We plotted the relative occupancy of the species (O i ) as a function of R i ,resultingin RSOC. Third, we compared which of the most common core–satellite species patterns (unimodal satellite dominant, bimodal symmetrical, or bimodal asymmetrical) gave the best fitfor the assemblages (Jenkins 2011). We fitted each of the following three SOFD patterns: 1. Unimodal satellite mode (exponential concave): O i =y 0 +a9exp(bR i ) where the initial parameters were y 0 = 0.01, a= 1.0, b= 0.01. 2. Bimodal symmetrical (sigmoidal symmetric): O i =a/(1 + exp(bR i +c), where the initial parameters were a=1.0,b=0.1, c=1.0. 3. Bimodal asymmetric (sigmoidal asymmetric): O i =a[1 exp(bR ic )], where the initial parameters were a= 1.0, b=1.0, c=1.0, where y 0 ,a,b, and care estimated parameters. The nonlinear regressions were used in the Levenberg–Marquardt algorithm (999 iterations) according to Jenkins (2011), and parameters were estimated by means of ordinary least squares (OLS) with IBM SPSS statistical package. We graphically evaluated the assumptions of the regressions for normality of residuals, homogeneity of variance, independent error terms, as well as the tails and shoulders of the data and models. We used Akaike information criterion for small sample sizes (AIC c )tocomparethealternative models. The model with the smallest AICc is considered to be best with respect to expected Kullback–Leibler information (Burnham and Anderson 2000). The approach is powerful to detect differences between models if DAICc ð¼ AICciAICcminÞvalues are higher than 4 (Anderson et al. 2000, Jenkins 2011). All the data analyses were performed using the IBM SPSS statistical package, version 23. RESULTS A total of 46 odonate species were recorded in the 292 waterbodies. On average, we found 8.4 [3.6 standard deviation (SD)] species, ranging from 1 to 18 (Table 3). In the combined data, the number of species did not increase with (log 10 - transformed) area of the waterbody (r=0.06, n=292, P=0.292). However, there were regional differences in the correlation between species number and waterbody area. In three out of the four regions, we found no such relationship (southern Finland, r=0.19, n=58, P=0.162; southern Sweden, r=0.10, n=94, P=0.353; and central Sweden, r=0.145, n=91, P=0.170). Only in central Finland, there was a clear positive relationship between waterbody area and species number (r=0.45, n=49, P=0.001). Each species occurred in an average of 54 62.4 (range: 1–209) waterbodies (Table 2). Overall, the species with a large geographical range occurred in a higher number of waterbodies (Fig. 2, Table 4). The model including breeding habitat and geographical range was considered the best of the tested models (Table 4). Generalists occurred in a larger number of waterbodies (mean: 96 74 SD, n=14) than specialists (39 47, n=28; generalized linear models, Wald =4.64, df =1, P=0.031; Fig. 2). However, the geographical range did not differ between breeding habitat types (Kruskal– Wallis test, H=5.37, df =2, P=0.068). Neither dispersal ability (measured by wing length) nor the geographical range of the species (r s =0.08, n=46, P=0.593) explained the species occupancy frequency in the waterbodies (Table 4). In the combined data, the SOFD pattern of the odonate species followed a unimodal satellite pattern (Table 5, Fig. 3). All alternative models ❖www.esajournals.org 6April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. fitted less well with data (DAICc >4; Table 5). There was a high number of satellite species, and half of the species (23 out of 46) occurred in <10% of the waterbodies. On the other hand, six species were found in at least half of the waterbodies (Fig. 3). We found geographical variation in the SOFD patterns (Table 5, Fig. 4). The three northern regions (i.e., central Finland, southern Finland, and central Sweden; Fig. 4a, b, d) showed the best fit with the bimodal core–satellite pattern, whereas southern Sweden followed the unimodal satellite-dominant pattern (Table 5, Fig. 4c). All alternative models fitted less well with data (DAICc >4; Table 5). More than half of the species occurred in <10% of the waterbodies in southern Sweden (Fig. 4c), but only about one-fourth in central Finland (Fig. 4b). Moreover, in Sweden, none of the species occurred in all waterbodies (Fig. 4c, d), whereas certain species, such as Aeshna grandis,A. juncea, and Coenagrion hastulatum, occurred in almost all of the studied Finnish lakes (Table 2, Fig. 4a, b). DISCUSSION Species richness We found a relatively high total number of odonate species in the studied waterbodies, about 77% (46 out of 60 species) of the total number occurring regularly in Sweden and Finland (Boudot and Kalkman 2015). We also found that species richness increased with waterbody area in central Finland, but not in southern Finland or in Sweden. The lack of a general relationship between waterbody area and species richness contrasts with the results of a previous study where larger ponds were shown to harbor a larger number of odonate species (Oertli et al. 2002). Larger lakes may have room for more niches, for example, different types/structures of aquatic plants, which increases odonate species richness (Oertli et al. 2002, Honkanen et al. 2011). However, the maximum waterbody size studied by Oertli et al. (2002) was 9.5 ha, whereas our waterbodies were much larger (up Table 3. The number of species found and mean, standard deviation (SD), and maximum (max) number and percentage of waterbodies occupied by odonate species in Finland and Sweden. Region Species Number of waterbodies Percent of waterbodies Mean SD Max Mean SD Max Southern Sweden 39 16.3 17.8 59 17.7 21.8 79.2 Central Sweden 38 20.3 22.5 75 21.1 23.5 73.1 Southern Finland 31 15.9 14.2 51 31.8 34.5 94.5 Central Finland 27 19.1 15.9 49 51.3 35.8 100.0 Combined 46 54.4 62.4 209 19.4 22.3 78.7 Geographical range area (1000 km2) 0 200 400 600 800 Number of waterbodies 0 50 100 150 200 250 Fig. 2. Number of waterbodies occupied by each odonate species (n=46) in relation to its geographical range in Finland and Sweden. Model prediction curve (continuous line) and 95% confidence intervals (dotted lines). The curve is based on the combined data set and calculated with generalized linear models. In the model, the number of lakes occupied by each odonate species was negatively binomial distributed with a logarithmic link function. The symbols denote the breeding habitat(s) of the species: Generalists (filled triangles) breed in both standing and running waters, specialists (open dots) breed in standing waters, and tourists (filled dots) breed mainly in running waters, but occasionally also in standing waters. ❖www.esajournals.org 7April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. to 247 ha). Moreover, diversity of many taxonomic groups are known to peak at high productivity (Dodson et al. 2000) rather than large area and previous studies have also noted that small lakes in forests often harbor a larger number of species than large ones (Flenner and Sahl en 2008, Koch et al. 2014). The fact that such a high percentage of the total number of recorded odonate species was found in our relatively small subset of Swedish and Finnish waterbodies is interesting, as the lakes surveyed by us constitute much less than 0.1% of all lakes in the two countries (estimated at >151,000 lakes larger than 0.01 km 2 ; Henriksen et al. 1998). This may be because the number of species encountered at these northern latitudes Table 4. Generalized linear models for the occupancy frequency of 46 odonate species in 292 waterbodies in Finland and Sweden. Model Parameter estimates Omnibus test AICc values IRange Generalist Specialist Wing G 2 df PAICc DAICc Range +Habitat 2.11 0.005 2.37 0.34 —48.39 3 <0.001 419.79 0.0 Range +Habitat +Wing 2.07 0.005 2.37 0.34 0.001 48.40 4 <0.001 422.31 2.52 Range 1.66 0.005 ———38.47 1 <0.001 425.02 5.23 Range +Wing 1.57 0.005 ——0.002 38.50 2 <0.001 427.28 7.49 Habitat 4.57 —2.60 0.90 —19.36 2 <0.001 446.42 26.63 Habitat +Wing 4.57 —2.97 0.90 0.006 19.51 3 <0.001 448.68 28.89 Wing 4.09 —— —0.004 0.06 1 0.815 463.43 43.64 Notes: Predictor variables were geographical range, breeding habitat (three categories: generalist, specialist, and tourist species as a reference category) and wing length of the species. Estimated parameters for the intercept (I) and predictor variables are shown in bold if they differed from zero (P<0.05).The adequacy of each model was tested by the goodness-of-fittest(G 2 ), and Akaike information criterion for small sample sizes (AICc) and DAICcð¼ AICciAICcminÞvalues are presented. The model with the lowest AIC c is considered as the best model of the tested. Em-dash indicates that the parameter(s) did not belong to the model. Table 5. Results of odonate species occupancy frequency distributions (SOFD) in Finland and Sweden. Region Figure Species AICc DAICc Combined 3 46 Unimodal satellite 327.1 0.0 Bimodal symmetric 322.8 4.3 Bimodal asymmetric 204.7 122.4 Southern Sweden 4c 39 Unimodal satellite 315.4 0.0 Bimodal symmetric 310.4 5.0 Bimodal asymmetric 178.3 137.0 Central Sweden 4d 38 Bimodal symmetric 272.9 0.0 Unimodal satellite 257.3 15.6 Bimodal asymmetric 154.7 118.1 Southern Finland 4a 31 Bimodal symmetric 272.9 0.0 Unimodal satellite 257.3 15.6 Bimodal asymmetric 154.7 118.1 Central Finland 4b 27 Bimodal symmetric 149.0 0.0 Bimodal asymmetric 130.5 18.6 Unimodal satellite 126.0 23.0 Notes: The three most likely SOFD patterns (unimodal satellite dominant, bimodal symmetrical, and bimodal asymmetrical) were analyzed with combined data and separately for the four different locations (southern Sweden, central Sweden, southern Finland, and central Finland). Figure column joins statistical models with data figures. Species denote number of species in each study region. Akaike information criterion for small sample sizes (AICc) and DAICc ð¼ AICciAICcminÞvalues are presented. The model with the lowest AIC c is considered as the best of the tested models. Proportion of waterbody area (%) 0 20 40 60 80 100 Number of species 0 5 10 15 20 25 Fig. 3. Number of odonate species (n=46) in relation to the proportion of the waterbody area occupied (%; n=292 lakes) in Finland and Sweden. ❖www.esajournals.org 8April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL. Valle, K. J. 1952. Die Verbteitungsverh€ altnisse der ostfennoskandischen Odonaten (Zur Kenntnis der Odonatenfauna Finnlands 6.). Acta Entomologica Fennica 10:1–87. Verberk, W. C. E. P., G. van der Velde, and H. Esselink. 2010. Explaining abundance-occupancy relationships in specialists and generalists: a case study on aquatic macroinvertebrates in standing waters. Journal of Animal Ecology 79:589–601. Watkinson, A. R., and W. J. Sutherland. 1995. Sources, sinks and pseudo-pinks. Journal of Animal Ecology 64:126–130. Watts, P. C., I. J. Saccheri, S. J. Kemp, and D. J. Thompson. 2007. Effective population sizes and migration rates in fragmented populations of an endangered insect (Coenagrion mercuriale: Odonata). Journal of Animal Ecology 76:790–800. Wikelski, M., D. Moskowitz, J. S. Adelman, J. Cochran, D. S. Wilcove, and M. L. May. 2006. Simple rules guide dragonfly migration. Biology Letters 2:325–329. Wittwer, T., G. Sahl en, and F. Suhling. 2010. Does one community shape the other? Dragonflies and fish in Swedish lakes. Insect Conservation and Diversity 3:124–133. ❖www.esajournals.org 15 April 2018 ❖Volume 9(4) ❖Article e02192 KORKEAM€ AKI ET AL.