Introduction

For most animals, mobility is a crucial trait that allows them to find basic resources such as mating partners, food or shelter. Moreover, mobility enables many animals to persist in increasingly fragmented and heavily altered environments in which more and more species are forced to live (Fahrig 2003; Foley et al. 2005; Concepcion et al. 2008). Due to such environmental changes individuals must be able to reach habitat patches with the essential resources in order to survive (Fahrig 2003; Thomas et al. 2004; Dover and Settele 2009). Within-patch mobility enhances local population resilience to the habitat disturbance and may also be an indication of the potential exchange of individuals between local populations which supports metapopulation persistence (Hanski and Ovaskainen 2000; Kraus et al. 2003). Individuals with large mobility may potentially explore all available resources within a habitat patch as well as utilise those that are widespread in different habitats (Dennis et al. 2003; Auckland et al. 2004; Kalarus et al. 2013).

Several insect studies revealed that flight performance is strongly associated with morphological traits (Berwaerts et al. 2002; Merckx and Van Dyck 2002). For instance, it was found that butterflies from newly established populations have longer wings (Merckx and Van Dyck 2002) and are characterized by increased mobility (Hanski et al. 2006). However, it remains unknown if morphological traits are also related to mobility at a very local, within-habitat patch scale.

Knowledge of mobility is especially important in the case of endangered species occurring in small and isolated local populations, which are surrounded by an inhospitable matrix (Mathias et al. 2001; Thomas et al. 2001; WallisDeVries 2004). One such species is a specialist butterfly, the dryad Minois dryas (Scopoli 1763). In Poland, the occurrence of this species is restricted to few isolated localities in the south of the country, including the Kraków region, where it survives in protected calcareous xerothermic grasslands, and more recently has spread into surrounding wet meadows (Dąbrowski 1999; Kalarus et al. 2013). Therefore, it is worth to examine whether the dryads living in wet meadows differ in mobility and flight morphology from those inhabiting xerothermic grasslands. It can be expected that the individuals from former group should be more mobile and cover longer distances, which should also be reflected in their morphology since they are the descendants of the most dispersive butterflies and their mobility is likely to possess a substantial genetic component (Hill et al. 1999; Merckx et al. 2003; Haag et al. 2005).

The aim of our study was to investigate differences in within-patch movements and flight morphology between the dryads from xerothermic and wet habitats. In addition, we focused on inter-sexual differences in mobility, which are typically strong in butterflies, but often neglected (Ovaskainen et al. 2008; Nowicki and Vrabec 2011; Schultz et al. 2012). Specifically, for both habitat types and sexes we compared (1) movement distances and turning angles; (2) morphological traits possibly associated with butterfly mobility (body mass, forewing length, thorax width, wing loading index, and general body size index); (3) resting times, which are inversely related to flight activity. Based on the results obtained we also provide practical conservation recommendations for the dryad.

Methods

Fieldwork

The survey was carried out during the flight period of the dryad, i.e. August, in 2009 and 2010. Four habitat patches were selected for the study. Two of them are xerothermic grasslands: Skołczanka (site A, 1.01 ha) and Uroczysko Kowadza (site B, 0.72 ha). The other two are nearby wet meadows (site C, 4.41 ha; site D, 1.04 ha). The inter-patch distances within the pairs of patches of the same type are 80 m (for wet meadows) to 180 m (for xerothermic grasslands), whereas both pairs lie 200–400 apart. All the investigated sites are located within the Bielańsko-Tyniecki Landscape Park and are part of the Natura 2000 site “Dębnicko-Tyniecki Obszar Łąkowy” (PLH 120065). A more detailed description of all four study sites can be found elsewhere (Kalarus et al. 2013).

We examined within-patch movements by the dryads, their resting times between consecutive flights, and morphological traits potentially related to mobility. The movements were recorded from the moment when a spotted butterfly was initially noticed resting on a plant. Butterflies were followed by observers from the distance of about 5 meters in order to avoid disturbing butterflies. Points where the butterfly stopped were subsequently marked with numbered bamboo poles. This made it possible to measure distances of movements and turning angles between them with the help of steel measuring tape and compass. Resting time between consecutive movements was recorded with the precision of one second. Other types of behaviour were very rarely observed and thus were not considered in the present study. After the fifth consecutive movement the butterfly was captured with an entomological net. Thorax width and forewing length were then measured to the nearest 0.1 mm using a Vernier calliper. Finally, each individual was placed in a small envelope and weighted with a TOMOPOL S-50 balance with the precision of 0.005 g. The weighting was repeated twice for each butterfly and averaged. Having taken all the above measurements, we marked the butterfly with a unique number on the underside of its hind wing with a non-toxic permanent marker, and immediately released it. The data from any individual (even if recaptured) were used only once in the analyses to avoid pseudoreplication.

Statistical analysis

Since our survey design was incomplete, i.e. not every site was surveyed in both years of the study; we were not able to test the effects of site and year separately. Thus, we instead adopted the effect of a data set, i.e. data collected at a given site in a particular year, as a random factor in the analyses conducted using generalized linear mixed models (GLMMs) with parameter estimates derived using the restricted maximum likelihood (REML) method (Searle et al. 1992; Wolfinger et al. 1994). The data set effect was always nested within habitat type. Nevertheless, in all the cases, it proved insignificant and did not improve the model fit. Consequently, with this effect removed, the final models were reduced to standard general linear models (GLMs).

Separate GLMs were applied to test the impacts of habitat type and sex as well as their interaction term on the morphological characteristics of the dryads. The investigated traits included body mass, thorax width, forewing length, wing loading index, and general body size (see below). The traits were derived for the purpose of subsequent analyses as the original traits were clearly positively correlated with one another (ln-transformed body mass vs. thorax width: r = 0.489, P < 0.001; ln-transformed body mass vs. wing length: r = 0.602, P < 0.001; thorax width vs. wing length: r = 0.589, P < 0.001). Wing loading index was defined as the ratio of body mass to forewing length. Applying the classic definition of wing loading (i.e. body mass to wing surface ratio; see Dudley and Srygley 1994) was not possible, because wing surface measurements could not be easily taken in the field, and thus we had to rely on a proxy using wing length instead (as e.g. Tiple et al. 2009). General body size index was calculated as the sum of standardised values of body mass, thorax width and wing length obtained through dividing the measurements by the means for all individuals. Body mass measurements were cubic rooted before standardisation. We decided to rely on the derived index rather than to use body mass as a measure of body size, because butterfly body mass varies (typically decreases) during adult life span, and definitely we were not able to weight all individuals in the same moment of their life, e.g. directly after hatching.

In the next step, further GLMs were built to explain patterns in the average movement distances and resting times. The latter was used as a measure of butterfly inactivity. Both variables were ln-transformed to achieve their normal distribution. Among the predictors the models included butterfly sex, habitat type and their interaction term as fixed effects; as well as wing loading and butterfly size indexes as continuous covariates.

Turning angles were grouped into twelve 30°-wide classes. Their distributions were first tested for inter-sexual differences separately for xerothermic grasslands and wet meadows with Chi square tests. Since there were no significant differences in any case, the data for both sexes were pooled together for further analyses of the turning angle distributions. The distributions were compared between the two habitat types, again using Chi square test. Additionally, the distributions for each habitat type were evaluated against two theoretical distributions: the uniform distribution and the von Mises distribution, which is equivalent to the normal distribution for circular data (Fisher and Best 1979; Batschelet 1981). The uniformity of the turning angle distributions was assessed with the Rao’s spacing test (Mardia and Jupp 2000), whereas the Watson’s U2 test (Zar 1998) was applied to examine whether the distributions were consistent with the von Mises distribution. We also calculated the concentration index K (Fisher and Best 1979) of the turning angle distribution for each habitat type to obtain how the distributions are concentrated around the mean angle. The concentration index K is a measure of the deviation of the angle distribution from a uniform one (Mardia and Jupp 2000). Higher K index values indicate higher concentration of angle distribution.

The GLMs were calculated in JMP 10, while the turning angle analysis was done in the Oriana 3.0 software.

Results

The GLM models for morphological traits indicated strong inter-sexual differences in each case (Table 1). Females of the dryad were heavier, had wider thorax and longer wings, but males had lower wing loading index (Fig. 1). Furthermore these inter-sexual differences were clearly confirmed for body size index in the interaction term habitat type × sex (Fig. 1). Individuals from xerothermic grasslands proved to have longer wings and lower wing loading index than those from wet meadows (Table 1; Fig. 1). In addition, the thorax was significantly wider in the former group, but the difference was evident only in males. For better illustration of morphometric measurements of the dryad butterflies their mean values are shown in Table 2.

Table 1 Results of the GLM model analysis of factors affecting flight-related morphological traits of the dryad
Full size table
Fig. 1
figure 1

Morphological traits (mean ± SE) of the dryad butterflies in both investigated habitat types. White and grey bars represent females and males respectively

Full size image
Table 2 Morphological traits and movement characteristics (mean ± SE) of the investigated dryad butterflies in relation to their sex and habitat type
Full size table

The analysis of distances covered by the dryads revealed a significant effect of sex, with males flying further distances than females, but not of habitat type (Tables 2, 3; Fig. 2). The majority of movement distances were quite short (>10 m) and very few exceeded 20 m, especially in females (Fig. 2).

Table 3 Results of the GLM model analysis of factors affecting average movement distances of the dryad butterflies
Full size table
Fig. 2
figure 2

Distribution of within-patch movement distances of the dryads. White and grey bars represent females (n = 75) and males (n = 52) respectively. To avoid psedoreplication a single movement per individual was randomly selected for the histogram

Full size image

In general, females tended to rest longer than males (Tables 2, 4). Although habitat type alone did not affect the resting times of butterflies, its interaction with sex was highly significant (Table 4). The GLM model results also implied a positive relationship between wing loading index and resting time (Table 4; Fig. 3). With this relationship accounted for, resting times were the longest in females from wet meadows and the shortest in males from the same habitat, while there was no particular difference between resting times of both sexes in xerothermic grasslands (Fig. 4).

Table 4 Results of the GLM model analysis of factors affecting average resting times of the dryad butterflies
Full size table
Fig. 3
figure 3

Relationship between wing loading index and resting time of the dryads

Full size image
Fig. 4
figure 4

Inter-sexual differences in resting times (mean ± SE) of the dryad in both investigated habitats. White and grey bars represent females and males respectively. To account for the significant relationship between butterfly wing loading index and resting time, the values presented are residuals of this relationship

Full size image

Distributions of turning angles did not differ between sexes in any habitat (xerothermic grasslands: χ2 = 4.251, P = 0.962; wet meadows: χ2 = 5.700, P = 0.893). However, with the data for both sexes pooled together, they did so between both habitats (χ2 = 21.281, P = 0.031; Fig. 5). Furthermore, the Rao’s spacing test outcome showed that the distributions were clearly not uniform, but instead they were consistent with the von Mises distribution as suggested by the insignificant Watson’s U2 test results (Table 5). In both habitat types the great majority of turning angles were centred near 0 (Fig. 5; see also relatively high values of the concentration index K in Table 5), which indicates that butterflies tended to move in fairly straight lines without much change in the flight direction.

Fig. 5
figure 5

Distribution of turning angles of the dryad flights within their respective habitat patches (a xerothermic grasslands, b wet meadows). The solid lines indicate the mean angle (presented with 95 % confidence interval). The inner scale shows frequencies of turning angles within 30°-wide classes

Full size image
Table 5 Comparisons of the dryad turning angles distributions in two investigated habitat types
Full size table

Discussion

Numerous earlier studies have documented the relationship between morphological traits of various butterfly species and their mobility. Specifically, it has been found that mobility is negatively affected by body mass, but positively by thorax width and wing length (Berwaerts et al. 2002; Skórka et al. 2013). The results of the present study generally confirm the above patterns, although we found that wing loading index is more useful as a factor explaining the dryad mobility. Wing loading index correlated positively, while body size index correlated negatively with resting time, but only the former relationship proved statistically significant.

Comparisons of morphological characteristics between the two investigated habitat types indicated that the dryads from xerothermic grasslands appear more adapted to flight. They had wider thoraxes and longer wings than the individuals from wet meadows, and consequently, lower wing loading index, since there was no difference in body mass between both groups. Nevertheless, despite apparently greater potential mobility of the butterflies inhabiting xerothermic grasslands, habitat type alone had no apparent impact on within-patch movement distances. However, these morphological differences may indicate that dryads from xerothermic grasslands are more capable to successfully migrate from natal patches than butterflies from wet meadows.

There was no clear effect of habitat type on the resting times. Females from wet meadows rested longer, but males rested shorter than the butterflies from xerothermic grasslands. In wet meadows, relatively long resting times in females may force males to take flight more frequently in order to search for mating partners. Taking everything into consideration, while habitat type indirectly affected the dryad mobility through its effect on flight-related morphology, we found no evidence for its additional direct influence on local mobility, although effects may be present on realized dispersal. Moreover, this indirect effect of habitat type was in the opposite direction than expected, i.e. the individuals representing recently established populations in wet meadows turned out to be less mobile.

Consequently, it seems justified to hypothesise that mobility and flight morphology of the dryad within the studied region is shaped by differences in resource availability among habitat patches as a result of phenotypic plasticity (cf. Merckx and Van Dyck 2006), rather than by the recent evolutionary history of the populations inhabiting them. It is known, that availability and quality of foodplants affect morphology of adult individuals (Fischer and Fiedler 2000). In earlier research on the dryad we have demonstrated that xerothermic grasslands offer substantially more nectar sources, whereas wet meadows are generally better in terms of larval foodplant abundance (Kalarus et al. 2013). Since nectar is the predominant source of energy utilised for flight by many butterfly species (Sacktor 1975; Kammer and Heinrich 1978), its increased availability supports highly mobile individuals. In contrast, the dryads living in wet meadows are likely to invest more in reproduction. It is worth noting that thinner thoraxes recorded in this habitat apparently imply larger abdomens, as body mass did not differ between both grassland types. Greater investment in reproduction presumably leads to less investment in flight capacity since a trade-off between both investments should be expected (Gibbs and Van Dyck 2010).

The flight distances of the dryads within their habitat patches were strikingly short in most cases. The movement distances recorded in our study were not only well below distances separating habitat patches, but also considerably shorter than patch dimensions. In addition, the fact that turning angles fitted well with the von Mises distribution indicates that the flight routine followed correlated random walk (Kareiva and Shigesada 1983). Altogether our findings suggest limitations to the dryad movements within their patches (cf. Hovestadt and Nowicki 2008).

Mobility depression in highly fragmented landscapes has been frequently reported in butterflies (Mathias et al. 2001; Hanski and Meyke 2005; Bergerot et al. 2012). The phenomenon has serious conservation implications as it decreases metapopulation viability (Hanski 1999; Russell et al. 2006). In our study system, the situation may be worsened by male-biased mobility. While females are often a more mobile sex in butterflies (e.g. Bergman and Landin 2002; Auckland et al. 2004; Nowicki and Vrabec 2011, but see Junker and Schmitt 2010), we observed the opposite pattern in the dryad. This is a further constraint for the successful conservation of this species because only mated females are able to ensure colonisations of vacant habitat patches, which is a key process for metapopulation persistence. It is also possible that low, male-biased mobility has led to the extinction of the dryad in northern Poland as was earlier reported (Romaniszyn and Schille 1929; Dąbrowski 1994); as well as to its decline in Germany (Settele et al. 1999).

Because the distances covered by the dryad are so low, the colonisation of empty patches may be difficult. The recommended conservation actions should be focused on enhancing the chances of continuous existence of local populations despite presumably low exchange of individuals. This should be achieved through maintaining relatively large habitat patches, which should prevent stochastic extinctions of their populations (Hanski 1999). In addition, the exchange of individuals among habitat patches may be promoted through improving permeability of the matrix. A good quality matrix supports butterfly dispersal as it encourages individuals to emigrate from their habitat patches (Ricketts et al. 2001; Schtickzelle and Baguette 2003). In the case of the dryad, the optimal type of matrix seems to be meadows lacking the foodplants, but rich in nectaring plants. However, as the dryad is also known to utilise sunny deciduous forests with sparse canopy and forest edges (Krzywicki 1982; Warecki and Sielezniew 2008), such habitats may also be useful for dispersal (cf. Streitberger et al. 2012). Finally, potentially poor colonising capabilities of the species should be an argument for its assisted reintroductions into vacant suitable habitats. Such actions spontaneously conducted in recent years, proved to be particularly successful in establishing new viable populations (Dąbrowski 1994).