Assessment of the quality of the terrestrial habitat of the threatened dragonfly, Sympetrum depressiusculum (Odonata: Libellulidae)

The majority of the conservation strategies for threatened dragonfl ies are designed to protect only their aquatic habitats. Sympetrum depressiusculum is a species threatened not only by the destruction of its aquatic habitats but also by its association with a specifi c terrestrial environment. In this study, we aimed to identify the key elements of the terrestrial environment of adult S. depressiusculum. We used generalized linear mixed models to determine habitat preferences of adults and the particular features of habitat patches, such as vegetation cover, vegetation structure and the availability of potential prey. Our results indicate that S. depressiusculum adults preferred mainly riparian vegetation but beyond ponds they utilized only certain terrestrial habitats (abandoned fi elds, meadows, forest clearings). Adults responded positively to habitat patches with a high cover of vegetation and suitable vegetation structure. Adult abundance was affected also by the distance of patches from the natal site. In an agricultural landscape, the availability of such habitat patches may be limited and could infl uence the abundance and distribution of this species. We suggest that conservation efforts for this species should not only focus on the larval environment but also include suitable surrounding terrestrial habitats. Effective management around natal sites should concentrate on maintaining a heterogeneous landscape, which is extensively managed (e.g. leaving several fi elds fallow, maintaining managed hay meadows).


INTRODUCTION
A suitable environment for an animal includes a mosaic of habitat patches that are suitable for all the activities essential for successful reproduction and population persistence (Orians & Wittenberger, 1991).It should be emphasized that the complex habitats of odonate species must meet these ecological requirements at all stages of their life cycle (Corbet, 1999).Moreover, responses to a stressful environment at any stage may carry over and shape fi tness in subsequent stages and generations (Stoks & Córdoba-Aguilar, 2012).
Adult dragonfl ies and damselfl ies in terrestrial environments use various habitat patches for mating, perching, foraging and nocturnal roosting (Corbet, 1999;Kirkton & Schultz, 2001;Rouquette & Thompson, 2007;Wildermuth, 2012).In addition, occurrence of adult odonates outside aquatic habitats may be associated with their dispersal (Suhling et al., 2015), and this may be infl uenced by landscape structure (e.g.Watts et al., 2004).The value of a high dispersal ability is important especially for those species occurring in patchy isolated habitats (e.g.ponds surrounded by a homogeneous agricultural landscape), mainly be-toma nasus (Linnaeus), which requires specifi c management, including periodic draining of the pond in late summer to promote growth of emergent vegetation.This management provides stable conditions similar to those in the natural habitats of S. depressiusculum.
Terrestrial habitats around the natal site consist of a mosaic of variously managed agricultural fi elds, meadows and forests.Agricultural areas (grains, rapeseed) and forests (natural softwood fl oodplain forests, oak-hornbeam woodlands, spruce monocultures) occupy the majority of the landscape area around the ponds.When sampling occurred some of the agricultural fi elds had been already harvested or were abandoned, while most of the meadows were managed less intensively.Abandoned fi elds were left fallow only for a short time and were in an early stage of succession and covered with common annual weeds, while areas abandoned for longer were usually covered with tall vegetation and shrubs.

Data collection
Habitat preferences of S. depressiusculum adults were recorded during the species' main fl ight period of July-September in 2014 and 2015.At the beginning of this research, we delineated patches of different habitats around the natal site (Fig. 1).During sampling, we repeatedly visited these patches and recorded the abundance of adult individuals.We counted adults between 10:00 and 16:00 CEST only on windless days when cloud cover was less than 20%.
In 2014, we monitored 38 patches at distances of 300 or 500 m from the natal site.The size of each patch was 10 × 10 m.For each patch, we recorded habitat (ruderal, arable fi eld, abandoned fi eld meadow, riparian vegetation, forest and forest clearing).In the middle of the season, when all the adults were fully coloured and sex was clearly determinable, the abundance of males and females was estimated separately.In 2015, we randomly selected 9 patches, all at a distance of 300 m from the natal site, where we repeatedly recorded adult abundance and several environmental factors: vegetation cover (total, tall vegetation, grass and dicotyledonous plants), height of the tall and ground vegetation (average of three random points in each patch), and abundance of potential prey up to 1.5 cm in size (mainly dipterans, but also other insects and arachnids).Arthropod abundance was sampled along three randomly placed 10 m transects in each patch using an insect sweep net (10 sweeps along each transect).Insects were generally identifi ed to order level.Table 1 provides a summary of the environmental factors measured in each of the patches and their ranges and medians.

Data analysis
We used generalized linear mixed models (GLMMs) with negative binomial error distribution and log link to determine the association between adult abundance (response variable), habitat and environmental factors (explanatory variables).In the GLMMs, environmental variables (habitat, distance from natal site, sex, arthropod abundance) were fi xed effects and patch visited was a mixed effect.For the analysis, we used the glmer.nbfunction in the lme4 package (Bates et al., 2014).For post-hoc comparisons between the mean number of adults in each habitat, we used the glht function in the multcomp package (Hothorn et al., 2015) with Tukey's pairwise multiple comparisons of means (Bretz et al., 2010).Statistical signifi cance was established using α = 0.05.Changes in abundance within a season were illustrated using generalized additive models and an integrated smoothness estimation from the mgcv package (Wood, 2016).All analyses were conducted in R 3.2.2 ( R Development Core Team, 2015).Conze, 2015).This species is associated with ephemeral habitats and spends a considerable part of its adult life cycle outside aquatic environments.Larval development lasts less than two months, while some adult individuals (the oldest specimen) live for almost 2 months (Raab et al., 2007;Dolný et al., 2014).Close association with terrestrial habitats in this species is therefore inevitably signifi cantly more important than for species associated with permanent habitats.The fi ndings of Dolný et al. (2014) indicate that the preference of S. depressiusculum adults for specifi c terrestrial habitats is not random, but determined by adults preferring only certain types of habitat, such as abandoned fi elds and small areas of ruderals.Unfortunately, that study does not convincingly determine which ecological factors affect the habitat preferences and use of terrestrial habitats by adult odonates' for their daily activities.Consequently, it is not known which specifi c management actions must be taken to protect such threatened species.Therefore, we present here the results of a study of adult preferences for specifi c types of terrestrial habitats.We hypothesized that the preference for particular terrestrial habitats is primarily determined by the availability of shelter and food, which are mainly associated with the type and structure of the vegetation.However, we must also include the effect of distance from the natal site, which could signifi cantly affect the number of adults fi nding these habitats.We therefore address three questions in this paper: (1) Which of the terrestrial habitats around wetlands are preferred by adults?(2) Does the distance and quality of suitable terrestrial habitats from the natal site affect the occurrence of adults?(3) Which habitat features infl uence the occurrence of adults in terrestrial habitats?

Study species
S. depressiusculum occurs in scattered populations from Central Europe to East Asia (Dijkstra & Lewington, 2006;Kalkman & Conze, 2015).Its natural water habitats are small, periodically fl ooded pools and lakes on fl oodplains of larger rivers.A key factor infl uencing the occurrence of this species is periodic drying out of aquatic environments (Verbeek, 1998;Schmidt, 2008).After the signifi cant loss of its natural habitats, this species switched to utilizing several types of artifi cial habitats, including fi sh ponds and rice fi elds that are dry during winter (Schmidt, 2008).Although this species is considered as endangered and very rare in Europe, it can be locally very abundant (Kalkman et al., 2010;Sternberg & Schmidt, 2000).In Central Europe, the larvae begin hatching in mid-April, and adult emergence starts at the beginning of July and continues until mid-August.When the weather is good the fl ight period ends in mid-October (Sternberg & Schmidt, 2000, Dolný et al., 2013).

Study sites
The study area is located in the foothills of the Beskydy Mountains (approximately 300 m a.s.l.) in the northeast of the Czech Republic.The surrounding landscape is primarily agricultural, with remnants of fl oodplain forest in the catchment area of the River Sedlnice.Larval development of S. depressiusculum was recorded at a single natal site (49°38´05.6˝N,18°06´04.3˝E),one of fi ve farm ponds situated close together.These pond have long been used for rearing the phytophagous cyprinid fi sh Chondros-

RESULTS
In 2014 and 2015, we recorded a total 2,022 S. depressiusculum adults.Habitat preferences of the adults were recorded throughout their occurrence, with the highest abundance in the patches recorded shortly after emergence and then decreasing throughout the season (Fig. 2).The GLMMs revealed that adult abundance was signifi cantly associated with habitat (χ 2 = 232.015,df = 6, P < 0.001) and distance from the natal site (χ 2 = 7.001, df = 1, P = 0.008).Adults were most abundant in patches with riparian vegetation.In a typical terrestrial environment, adults occurred mainly in abandoned fi elds, meadows and forest clearings.Imagoes avoided fi elds with crops and those covered with ruderals and no adults were recorded in forests (Fig. 3).Habitat preferences of adults were not affected by the distance of the patches from the natal site (habitat-distance interaction: χ 2 = 6.940, df = 6, P = 0.326), but their abundance was signifi cantly lower in the more remote patches.
During sampling in 2014, we recorded a total of 542 males and 172 females.Distribution of adult males and females was also associated with habitat (χ 2 = 62.5, df = 1, P < 0.001), but males and females did not differ in their preferences for particular habitats (sex-habitat type interaction: χ 2 = 11.7,df = 6, P = 0.069).Moreover, the effect of distance from natal site on the abundance of adult males and females did not differ signifi cantly (χ 2 = 3.451, df = 1, P = 0.063).
Adult abundance was not associated with the location of individual patches in the area (e.g. the presence of ponds near the patch) (Fig. 4).In other words, the occurrence of adults was affected primarily by the patches' specifi c microhabitat conditions.The number of adults was significantly affected by the total cover of vegetation (χ 2 = 7.862, df = 1, P = 0.005), cover of tall vegetation (χ 2 = 17.215, df = 1, P < 0.001) and abundance of arthropods (χ 2 = 32.59,df = 1, P < 0.001).The other factors associated with microhabitats (especially vegetation) had no signifi cant effects.In addition, the abundance of the species studied was not associated with the abundance of any particular insect order.

DISCUSSION
Based on our results, we conclude that adults of S. depressiusculum prefer terrestrial patches with specifi c properties.Adults respond positively to habitats with a high cover of vegetation and suitable vegetation structure.Such features are only associated with certain terrestrial habitats such as abandoned fi elds, extensively used meadows, some areas with ruderals or forest clearings.In intensively farmed agricultural landscapes the availability of such patches may be limited and could consequently infl uence the abundance and distribution of this species.From a life cycle perspective, the quality of terrestrial environments can therefore be as important as that of the aquatic environment necessary for larval development.
Adult individuals were associated especially with riparian vegetation, but away from ponds they utilized only a fraction of the available terrestrial habitats.High adult density in a riparian zone is a common phenomenon in dragonfl ies because most activities associated with their reproduction take place in the immediate surroundings of aquatic habitats (e.g.Corbet, 1999;Suhling et al., 2015).Adults used riparian vegetation growing along the embankments of intensively managed ponds (not their natal site) for pairing and copulation.Littoral zone in the vicinity of this vegetation was also used for oviposition (pers.obs.).However, these ponds are not suitable for larval development and could be regarded as ecological traps (see Šigutová et al., 2015).In the surrounding terrestrial landscape, adults preferred abandoned fi elds, meadows and forest clearings and clearly avoided arable fi elds, forests, and some areas with ruderals.This is similar to the results of Dolný et al. (2014), who note that even patches of terrestrial habitats   within an agricultural matrix could constitute a mosaic of preferred and non-preferred habitats for dragonfl ies.
Preference of S. depressiusculum for specifi c terrestrial habitats could be associated especially with vegetation structure.It is unlikely that adult odonates are associated with particular species of plants, as are some pollinators and herbivores, but more likely they respond to such specifi c vegetation properties as structure and density (Buchwald, 1992;Foote & Hornung, 2005).Adults of S. depressiusculum responded positively to the cover of tall vegetation but not to the height of the vegetation, as was expected.Dense vegetation provides suitable shelters and opportunities for foraging (Sternberg & Buchwald, 2000;Wildermuth, 2012).In addition, tall vegetation was used primarily as perching sites (pers.obs.), which is typical of the sit-and-wait predators of the genus Sympetrum (Corbet & May, 2008).Within the area studied, habitats with a suitable vegetation structure were abandoned fi elds, extensively managed meadows and forest clearings.Homogeneous patches without structured vegetation, which are typically avoided by adults, included arable fi elds, intensively managed hay meadows and some areas with ruderals.Large differences in the adult abundance recorded in different patches of ruderals were apparently associated with their variability in vegetation structure, as several patches of ruderals were colonized by an invasive plant (e.g.Solidago sp.), which formed dense, unstructured vegetation.
Although adult dragonfl ies' generally forage more frequently in areas with a higher abundance of potential prey (Corbet, 1999;Kirkton & Schultz, 2001), there was no association between the abundance of adults recorded in this study and the occurrence of any insect order, such as dipterans, which are considered to be the main prey of dragonfl ies (Corbet, 1999;Olberg et al., 2000).This indicates that the positive correlation between the abundance of S. depressiusculum and all arthropods could be mainly due to both groups preferring more structured vegetation (e.g.Zahn et al., 2010).It is likely, therefore, that the availability of prey in certain habitats does not primarily affect the habitat preferences of the species studied.However, prey density may be signifi cant, especially for territorial species and other species that utilize only the habitats occurring in the immediate vicinity of an aquatic habitat.

Management implications
Currently, S. depressiusculum is one of the most endangered species of dragonfl y occurring in Central Europe.It is also worrying that the majority of the Central European populations of this species occur in artifi cial aquatic habitats, such as fi shponds (Kalkman et al., 2010).Moreover, some of these habitats are situated in intensively managed, homogeneous agricultural landscapes.We suggest that terrestrial habitats could be as important as aquatic habitats for population viability and consequently efforts to conserve this species should not only focus on the larval environment but also consider terrestrial habitats surrounding the natal site.Our results indicate that for their routine activities S. depressiusculum adults utilize primarily habitats with heterogeneously structured vegetation.It is evident from our fi ndings that effective conservation management should concentrate on maintaining a heterogeneous landscape such as by leaving several fi elds fallow, use of extensively managed grazing, or maintenance of extensively managed hay meadows (mowed twice annually).Such land use would provide a suitable vegetation structure for this dragonfl y.
It is widely assumed that adult odonates avoid recently mown habitats (Sternberg & Sternberg, 2004;Wildermuth, 2012;Dolný et al., 2014), but very little is known about the direct effect on adult mortality of harvesting.As S. depressiusculum adults are relatively weak fl iers (Sternberg & Schmidt, 2000;Dolný & Holuša, 2008) and individuals aggregate in certain patches, mowing or harvesting at the wrong time (e.g. during cloudy or cold weather when adults are inactive) could signifi cantly increase mortality.Therefore, the fi rst mowing of meadows should be done before the emergence of adults (at the beginning of July) and the last at the end of their occurrence (mid-September).An alternative is to maintain mosaic-like mowing regimes.

Fig. 1 .
Fig. 1.Location of the patches studied.Patches marked with an asterisk were monitored in 2015.

Fig. 3 .
Fig. 3. Abundance of adults recorded in each habitat.Mea -meadow; ForC -forest clearing; RiV -riparian vegetation; AbF -abandoned fi eld; ArF -arable fi eld; Rud -area with ruderals; For -forest.Letters above abbreviations indicate signifi cant differences in the adult abundances recorded in the different habitats based on a post-hoc Tukey's pairwise multiple comparisons of means.

Fig. 4 .
Fig. 4. Effect of patch location on abundance of adults in 2015.Circle size indicates the relative average abundance of adults in each patch.The natal site is located in the middle (A).

Table 1 .
Summary of environmental factors measured.