Within-habitat vegetation structure and adult activity patterns of the declining butterfly Euphydryas aurinia

Euphydryas aurinia is a declining butterfly inhabiting oligotrophic grasslands in Central and Western Europe. Despite numerous ecological studies, patterns of its adult activity have so far been rather neglected, although adult resource use contributes to resource-based understanding of insects’ habitats. To relate E. aurinia adult activity patterns to within-habitat vegetation structures. (1) Timed adult activity observations along a transect crossing a colony site, analysed via partial ordination methods. (2) Activity records obtained during mark-recapture, analysed via binomial regressions. Both methods, besides influences of weather, time of day (similarities between morning and late afternoon hours), and progression of season (mate locating replaced by maintenance activities), revealed consistent association of behaviours to vegetation structures. Of the two male mate-locating behaviours, perching occurred near shrubs and woodland edges, and patrolling over centres of inhabited meadows. Female activity concentrated in nectar-rich mid-height sward near host plants. Consequently, male and female activity were partly spatially separated. A habitat for E. aurinia should provide resources for all its activities in close proximity. Grasslands containing host plants should be dissected by structures such as shrubs, woodlot edges, or taller herbaceous vegetation, emphasising the importance of landscape heterogeneity for insect fauna.


Introduction
Preserving suitable habitats is the crucial condition for efficient conservation of invertebrates (Samways 2007) including butterflies (Settele et al. 2009;Warren et al. 2021). Increasingly detailed information on habitat requirements of increasing numbers of specialist butterflies (e.g. Kivinen et al. 2008;Turlure et al. 2011;Maes et al. 2014;Vrba et al. 2021) indicate that a habitat patch must contain all vital resources for diverse activities of the given species, including larval host plant and shelter, overwintering substrate, adult food and shelter, and substrates utilised for mate locating and copulation (Dennis 2010;Turlure et al. 2019). Such resources may exist syntopically (cf. Courtney and Duggan 1983) or disjunctly (cf. McKay 1991) in time and space, but they must be present within routine individual movements' distances of the species/developmental stages concerned (Dennis 2010). Consequently, the habitat of a butterfly species should not be viewed merely as a "vegetation community" containing the species' larval host plant, or as a land cover category recognised, e.g., by geographers or landscape planners (Vanreusel and Van Dyck 2007). This distinction is crucial for managing reserves, biological restoration, and recovery programmes. For instance, the finding that behavioural responses to vegetation structure affect the reproductive fitness of individuals, quite accepted in vertebrate conservation (Caro 2007), implies that managing vegetation structure may enhance, or suppress, local populations of rare and endangered species (Shreeve and Dennis 2011;Turlure et al. 2011).
Here, we explore the relations between vegetation structure and adult E. aurinia activity on humid grasslands in Western Bohemia, Czech Republic. For several years, we surveyed the population using mark-recapture (Zimmermann et al. 2011a, b) supplemented by other methods targeting the species' conservation requirements. This paper analyses two types of evidence: (a) observation of the butterfly activity, obtained during repeated timed walks along a fixed transect crossing a colony site; and (b) records of behaviour and substrate prior to the captures, recorded during marking the butterfly. Using these two approaches, we investigate adult temporal behavioural patterns and relationships between adult activity and structural features of its habitat. We predict that distinct behaviours will be associated with distinct habitat structures and that the association will be detectable despite activity changes due to weather conditions, time of day, and progressing season.

Study system
Field data originated from a network of humid seminatural grasslands near Karlovy Vary, Western Czech Republic (50° 9′ N, 13° 2′ E, altitude 650 m), on a hilly piedmont of the volcanic Doupovské Mts. We worked within a network of 28 ha humid meadow patches, separated one from another by ponds, shrubby hedges, and woodlots. In terms of vegetation, they are classified as intermittently wet Molinia meadows (association Junco effusi-Molinietum caeruleae Tüxen 1954; Fig. 1). The wider area represents a regional stronghold for humid grassland butterflies (Fric et al. 2010).
The population of Euphydryas aurinia develops solely on Succisa pratensis (Moench, 1794) on which the larvae feed gregariously until hibernation and solitarily in the spring. The adults emerge in late May and the emergence is protandrous, ♀♀ appear a few days after ♂♂. The flight period lasts about three weeks (Zimmermann et al. 2011a). While ♀♀ split their time between nectaring and egg laying, ♂♂ invest much time into mate-locating activities. Notably, they use two distinct mate-locating tactics, perching and patrolling (Wahlberg 2000;Wahlberg et al. 2001). The same ♂♂ individuals can alternate these two activities during their lifetime (unpublished data).

Transect observation of adult activity
In 2003, concurrently with marking the butterflies (carried out 28 May-17 June, cf. Zimmermann et al. 2011a), we set a fixed transect (total length: 970 m), crossing the meadows inhabited by E. aurinia. It was divided into 15 sections (mean length: 63 m ± 32.2 SD, range 20-150 m), separated 1 3 by distinct landmarks and characterised by vegetation physiognomy (Fig. 1).
For eight days spread across the flight period (30-31 May; 2, 5, 7-10 June), we repeatedly walked the transect between 8:30 a.m. and 17:30 p.m. (CEST), usually twice per hour but with some variation due to weather, summing up to 132 walks (mean per day: 16 ± 1.3 SD, range 12-18). During the walks, we recorded all E. aurinia individuals seen per walk and section as in standard Pollard transects (Pollard and Yates 1993) but considering a smaller distance to the recorder (an imaginary 3 × 3 × 3 m cube), so that it was possible to record their behaviour.
The behaviours recorded, modified to fit the situation in E. aurinia, were, in ♂♂: Basking, Flight (direct, uninterrupted), Patrolling (gliding low over vegetation), Perching (settled at a prominent landmark, such as a tall grass blade or overhanging tree branch, at a sunny spot, ready for take-off), Chasing (with another butterfly/ insect), Mating, Nectaring, and Resting (wings closed, hidden position, usually in shade). In ♀♀, we recognised Basking, Flight, Oviposition, Chasing, Mating, Nectaring, and Resting.
We analysed the data using canonical correspondence analysis (CCA), a unimodal ordination technique relating the composition of samples to external predictors, in CANOCO v.5 (Ter Braak and Smilauer 2012). Each section walk represented a sample, and the activity records were the multivariate dependent variables. Significances of the ordinations were tested using 999 Monte-Carlo permutations, accounting for the spatial (consecutive sections of the transect) and temporal (repeated walks) structure in the data by using a split-plot design, permuting the data as cyclic shifts on both whole-plot and split-plot levels. Because this permutation design does not allow "empty" samples, a small number (0.001) was added to each cell in the response data table.
Targeting the response of activities to vegetation, we first controlled for nuisance effects of transect length, hour, and serial day (cf. Vlasanek et al. 2018). We selected the bestfitting response to hour and day from linear, polynomial, and factorial codings, and used the coding explaining the highest amount of variation in response data (Var%). Next, we assessed the response to weather. Finally, we constructed a covariate model containing hours, day, and weather variables selected by forward selection, and vegetation variables as predictors.

Activity records from mark-recapture
MR data originated from 2002 (24 May-28 June, 1141 behavioural records), 2003 (28 May-17 June, 2852 records), and 2004 (27 May-15 July, 2642 records). The marking was realized in a standard way: the butterflies were netted, marked with unique codes using alcohol-based pens, and released at points of capture. For every handling event, we also recorded the individual's sex, time of capture (closest hour), weather (Sky and Wind, using the same system in the transect walks), and the butterfly's behaviour prior to capture.
As above, we distinguished Resting, Basking, Nectaring, Flight, Patrolling, Reproduction (mating in ♂♂, mating plus oviposition in ♀♀), and Chasing. In addition, we recorded if the activity occurred at meadow edge (two-level categorical predictor), defined as located within approximately 3 m distance perpendicularly from a contiguous vertical wall of trees or shrubs, and near a host plant (again two categories, yes or no), again defined as approximately 3 m apart from the closest host plant. Edge data were recorded in all three years, host plant data only in 2003 and 2004. Regression models relating the occurrence of the activity patterns to habitat edge or host plant were constructed using the glm function with binomial distribution in R 3.6.2 (package stats, family "binomial", link function "logit"). For each type of behaviour, separately for ♂♂ and ♀♀ and for each of the three years, the modelling followed the same routine. We first entered variables describing weather during the individual observation, influencing butterfly activity considerably and rapidly. Second, we entered the hour in linear or quadratic forms to account for systematic effects of diurnal activity rhythms. Third, we entered the effect of serial day, again in linear and quadratic form, to account for possible seasonal effects. We used ΔAIC (≤ 2.0) to decide which variables to retain in the respective model. To the thus constructed covariate models, we sequentially added edge (all three years) and host plant (2003 and 2004 only) effects, again using AIC-statistics to decide whether either of these two predictors, or their combination, improved the fit of the model in question.
In the CCA analyses, all potential covariables influenced the distribution of records significantly (Table 1), although there were notable differences. Section length was a weak predictor, indicating that other circumstances were much more important. Section identity, in contrast, was the strongest of all predictors, clearly because the sections differed in vegetation structures, and butterfly activity reflected this. For weather (Fig. 2a), the distinction Sun plus Sky versus Dew represented the main gradient. In both sexes, basking, resting, and (less so) nectaring were positively associated with wet sward, low Sun and overcast Sky, while the opposite applied for chasing, ♂♂ patrolling and perching. On the secondary ordination gradient, resting of both sexes was often observed in windy conditions. Hour fitted the data best if coded as a category (Fig. 2b). The main gradient of variation was between mornings plus late afternoons, when both sexes mainly rested or nectared; and mid-days, when chasing, ♀♀ oviposition, and ♂♂ patrolling and perching peaked. Mating peaked in the afternoon hours. Ordination with day ( Fig. 2c), again best coded as factor, revealed a difference between early and late flight period, with ♂♂ prevailing in early and ♀♀ in late season. For ♂♂, the peak of patrolling seasonally preceded those of perching and nectaring; while for ♀♀, the peak of mating preceded that of nectaring and oviposition.
Vegetation alone explained second-highest proportion of variation in the behavioural records after section identity, and retained its significant effect if controlled for hour, day, and weather, and even section, and in a complex covariate model (Table 1). The ordination diagrams, both without control for covariates (not shown) and after filtering their effects (Fig. 2d) showed that the first ordination gradient distinguished between mid-to tall sward with rich nectar and a high host plant representation, and conditions with shrubs and short sward. The tertiary gradient (not shown) distinguished short sward from mid-and tall sward. ♀♀ oviposited, basked, nectared, and rested near the host plants, though resting occurred in taller sward than the other activities. ♂♂ rested most frequently in tall sward near shrubs, perched and chased other insects near shrubs at shorter sward, and patrolled, nectared, and basked at short to medium sward, independently of host plant abundance. It follows that some ♂♂ activities, especially those associated with mate locating, were spatially separated from spots with high host plant cover, where ♀♀ performed most of their activities.
Results of the binomial regressions (Tables 2, 3), despite some inconsistencies among years, agreed with the established knowledge of adult butterfly activities. For instance, ♂♂ basking decreased with sunny weather (all three years), ♀♀ mating decreased (2003,2004) and nectaring increased (♂♂: 2003 and 2004, ♀♀: 2004) with flight period duration, and ♂♂ perching followed domed patterns, indicating peaks in the middle of the adult period (all 3 years).
Regarding the within-habitat structures, the results were consistent across the three years for ♂♂ perching and patrolling. Perching prevailed near edges in all three years, plus outside of host plant patches in 2003. Patrolling consistently prevailed over centres of

Discussion
Using two alternative approaches, timed observation of activities along a fixed transect and analysis of capture circumstances during mark-recapture, we linked withinhabitat vegetation structures to adult activity patterns of the Euphydryas aurinia butterfly. Out of two mate-locating behaviours of ♂♂, perching was closely associated with meadow edges, i.e., with trees, woodland margins, lines of shrubs, or just unmanaged tall herbaceous vegetation surrounding regularly mown grasslands containing the host showing Euphydryas aurinia activity records collected along a fixed transect route across a colony site (cf. Fig. 1   Both our approaches produced complementary results and both withstood robust controls for short-term weather conditions and temporal aspects of adult flight. While the behavioural patterns due to weather (resting in windy overcast conditions, basking when the sward was wet by morning dew or after rains, behaviours associated with movements in sunny conditions) were rather trivial, the diurnal and the seasonal patterns displayed clear structuring. Interestingly, morning and late afternoon hours shared many similarities. The prevailing activities in 8:00-10:00 in the morning and 16:00-17:00 in the afternoon were basking, nectaring, and resting. This was arguably connected to the intake of energy, both thermal (cf. Franzen et al. 2022)    and nutritional (cf. Botham et al. 2011) for commencing daily activities in the mornings, and to replenish the energy before night rest in the evenings (cf. Vlasanek et al. 2018). The energy demanding activities such as male perching (with frequent conspecific and heterospecific chases) and patrolling, and female oviposition, culminated in 11:00-14:00, i.e., around noon. Franzen et al. (2022) described peaks of E. aurinia activity in early afternoons, attributing it to the thermal requirements of this early-season species, and showed that activity tends to decline under extremely hot temperatures. Whereas diel activity patterns displayed a similarity between mornings and evenings, seasonal patterns were linear, reflecting the changing proportion of sexes and changing status of the individuals along the flight period. Perching and patrolling peaked early in season, when ♂♂ prevailed in the studied colony (Zimmermann et al. 2011a) and were later followed by the peak of "maintenance activities", especially nectaring. Vlasanek et al. (2018) observed this pattern for multiple temperate species with distinct generations. Notably, perching and patrolling were not segregated temporarily. Regarding ♀♀, besides later seasonal peaks of all activities attributable to protandry (Schtizckzelle et al. 2005;Zimmermann et al. 2011a), the ordination analysis showed that oviposition and nectaring were postponed relative to mating. A female must evidently have mated to oviposit, and intake of nectar presumably increases the number of egg batches produced (O'Brien et al. 2004).
The relationships between behaviours and vegetation structures remained apparent after considering the above weather and temporal effects. The most robust patterns concerned mate-locating activities. The association of perching with shrubs near short sward (ordination) and with edges (regressions) conforms with the classic (Scott 1974;Dennis and Shreeve 1998;Rutowski 1991;Wickman 1992) concept of perches as conspicuous landmarks in windshielded locations provided by edges of taller vegetation, but with good oversight of the habitat. Perches may occur at host plant patches (as in the skipper Carterocephalus palaemon (Pallas, 1771): Ravenscroft 1994, or the satyrine Coenonympha pamphilus (Linnaeus, 1758) : Wickman 1985) or at patches of other critical resources, notably nectar (as in the copper Lycaena hippothoe (Linnaeus, 1761): Fischer and Fiedler 2001;Turlure and Van Dyck 2009). In our case, both the ordination and regressions suggested that the location of perches was independent of host plants distribution. Patrolling, on the other hand, occurs over midsward with a good supply of nectar, in the central parts of individual meadows (regressions) and rather close to host plants (ordination). Such a setting allows patrolling males to spot freshly hatched virgin females and at the same time, the unmated females to approach perching males. The latter consideration may deserve further attention, because Pinzari et al. (2019) observed that a fraction of females of Euphydryas aurinia provincials (Boisduval, 1828) mate a few days after emergence. The location of perches away from host plants indicates that perching males unlikely harass females seeking for oviposition, a situation described for L. hippothoe by Turlure and Van Dyck (2009), but does not exclude potential harassment of females by patrolling males.
The whole situation with perches situated near edges independently of host plants distribution, and males alternating perching with patrolling over host plants patches suggests that an ideal E. aurinia habitat would be finely structured, with shorter and taller swards, leeward edges, and host plant patches alternating at small scales of routine within-habitat movements. In such a setting, host plants with increased activity of females would naturally occur near edges with male perches. In traditional landscapes (cf. Loos et al. 2014;Perovic et al. 2015), and presumably in naturally patchy landscapes of prehistory (cf. Fahrig 2017), proximity of open short sward patches and edges of all types was probably a rule, and restoring highly heterogeneous conditions, e.g., by forest grazing (Saarinen et al. 2005), could boost E. aurinia populations. Several authors across the E. aurinia range disclosed "heterogeneity" as a positive correlate of its presence and density (Munguira et al. 1997;Scherer and Fartmann 2021), while others warned against uniformising management (Johansson et al. 2019). It is not unlikely that mate-seeking ♂♂, in contrast to ovipositing ♀♀, cannot recognise host plant patches, which may occasionally cause establishment of perches too far from females' activity; such a situation was described in a patrolling butterfly, Parnassius mnemosyne (Linnaeus, 1758) (Konvička et al. 2007).
From a more general insect conservation perspective, detailed understanding of E. aurinia adult activity provides further evidence that large uniformly managed land units exceeding in size routine within-habitat movements (cf. Baguette and Van Dyck 2007), even if containing abundance of some resources (e.g., larval host plants), will always be inferior to patchily heterogeneous environment (cf. Liu et al. 2006). This knowledge, increasingly accepted by the conservation community over the past decades (Rundlöf and Smith 2006;Kivinen et al. 2008;Lebeau et al. 2015;Perovic et al. 2015;Schwarz and Fartmann 2022) gradually transferred into reserve management on small scales, including our study system. The practices applied to conserve grassland insects may include patchy mowing, retention of uncut fallows, or grazing via small panels. Regretfully, diversifying grasslands' management remains difficult at large scales and beyond protected areas, including the matrix separating E. aurinia occupied sites in the western Czech Republic (Junker et al. 2021), although evidence is accumulating that much can be gained by relatively cheap measures, such as dissecting vast management units by temporary fallows or hedgerows (Buri et al. 2013;Bruppacher et al. 2016;Salek et al. 2018).

Acknowledgements
We are grateful to the Czech Science Foundation (P505/10/2167), South Bohemian University (144/2010/100), and the Technology Agency of the Czech Republic (SS01010526) for financial support. Matthew Sweney corrected the English.
Author contributions All authors conceptualised and designed the study and all of them equally participated in collecting the data in the field. KZ and MK curated the data in time between the fieldwork and manuscript preparation. MK, PV and ZFF performed the analyses; ZFF prepared the artworks, MK and ZFF wrote the article. All authors reviewed the manuscript.
Funding Open access publishing supported by the National Technical Library in Prague.

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.