Introduction

Over recent decades, butterflies have become pivotal indicators in ecology and environmental adaptation studies (Warren et al. 2021; Sunde et al. 2023). Often associated with sunny, transient habitats, butterflies are vulnerable to climatic and land-use shifts, especially in northern Europe’s fragmented landscapes (Kindvall et al. 2022). Many butterfly species are declining primarily due to habitat degradation, emphasising the need for robust conservation strategies (Warren et al. 2021). Protecting species in semi-natural grasslands becomes challenging when reliant on age-old land-use practices and threatened by intensive livestock grazing (Ellis et al. 2012; Kindvall et al. 2022). Many butterflies exist within metapopulations, distinct units interlinked by dispersal (Hanski 1998; Saccheri et al. 1998; Ranius et al. 2011). Their constant movement, driven more by daily needs than intrinsic dispersal motives, underscores the nuanced nature of their behaviour (van Dyck and Baguette 2005). Current understanding hinges on studies of fragmented populations, including isolated and migratory species (Shreeve 1995; Hanski et al. 2011). However, regions with abundant suitable habitats often remain under-researched due to the challenges they present, such as the difficulty in delineating habitats, the huge resources required to study large and mobile populations over time, problems in separating routine movements from dispersal, and the complexity involved in designing and conducting studies that can effectively capture the spatiotemporal dynamics of these ecosystems. We must also focus on more widespread areas sustaining endangered species and populations to enhance our understanding of biodiversity dynamics.

Existing literature has revealed multiple factors influencing butterfly habitat selection and movement, including individual attributes (e.g., age and sex), habitat features such as nectar availability and vegetation, and the spatial configuration of the landscape (Warren 1987; Wiens 1989; Hsiung et al. 2018). Increasing studies have focused on animal movements in relation to land cover (Loos et al. 2015; Brown et al. 2017), and the findings offer landscape ecologists insights into animal perceptions of environmental variance and the resulting macro-level effects (Turlure et al. 2011). Additionally, the density of conspecifics has been identified as important, often influencing dispersal (Enfjäll and Leimar 2005). While previous studies offer critical insights into these relationships, they often neglect the potential differential impacts of land cover variables on patterns of departures from and arrivals in sites (Kareiva et al. 1990; Baguette et al. 1998; Cowley et al. 2001; Ims and Andreassen 2005). Understanding the interplay between land cover, population densities and the movement dynamics of endangered butterflies is crucial for devising sound conservation strategies. Therefore, this study’s focus on dissecting these relationships is not merely additive to the field; it is pivotal for crafting more effective, evidence-based conservation measures that could determine the survival of endangered species within rapidly changing ecosystems.

In this study, we offer an in-depth examination and comparison of movement patterns in three declining and locally red listed butterfly species (Maes et al. 2019): Marsh Fritillary (Euphydryas aurinia, Rottemburg, 1775; Nymphalidae), Apollo (Parnassius apollo, Linneaus, 1758; Papilionidae), and Large Blue (Maculinea/Phengaris arion Linneaus, 1758; Lycaenidae). Our study area in the northern part of the island of Gotland in the Baltic Sea is known for its extensive suitable habitats, which support some of the world’s highest population densities for these species, at least in favourable years (Franzén et al. 2024; Sunde et al. 2024). This unique setting provides an unparalleled opportunity to study the dynamics of these butterfly species in conditions where their populations can thrive.

Understanding the movement patterns of butterflies has long been a cornerstone of conservation biology and landscape ecology, reflecting broader concerns about biodiversity, habitat fragmentation, and ecological networks (Baguette et al. 2013). Research has traditionally distinguished between two primary types of movements: routine, localised movements within habitat patches, often driven by immediate needs such as foraging, mating, and thermoregulation. (Stevens et al. 2010), and dispersal movements, which involve longer distances and are crucial for gene flow, population dynamics, and the colonisation of new habitats (Clobert et al. 2001; Bowler and Benton 2005). While much of the butterfly movement ecology literature focuses on the former, emphasising daily activities within specific habitat patches (Warren et al. 2001) (Dover and Settele 2009), there is a growing recognition of the need to better understand the triggers and consequences of dispersal movements. These are especially relevant in the context of environmental change and habitat fragmentation, which increasingly challenge the persistence of butterfly populations (Schtickzelle et al. 2006).

Our study aims to bridge this gap by examining butterflies' departure and arrival events across hectare grids, moving beyond the confines of localised movements to explore the factors influencing dispersal and longer-distance movements. Specifically, we investigate if land cover properties trigger butterflies to leave a particular one-hectare grid and arrive in another. Our core aim was to discern whether departures from and arrivals in hectare grids of these species correlate with local population density and land cover variables, including open vegetated land, ground moisture index, and forest cover. To this end, we assessed the relationships between land cover, hectare grid-level population density, and whether the proportion of butterflies departing and arriving in these hectare grids differed from grids where they were previously recorded.

Materials and methods

Description of the study area

Our study encompasses a 60 km2 region near Slite on Gotland Island, Sweden (midpoint: 57°69′ N, 18°69′ E) (Fig. 1). This area, one of Europe’s last refuges for these species, is distinguished by its calcareous bedrocks and vast natural habitats. These support sustainable populations of the butterfly species we investigated, including the Marsh Fritillary, Apollo, and Large Blue. So far, these geological and ecological features have created a unique sanctuary for these species, offering them a stronghold against the global trend of habitat degradation and population declines (Franzén et al. 2024). The climate is temperate, with cool summers averaging 16.6 °C in July, cold winters dipping to − 2.1 °C in February, and an average annual precipitation of 524 mm (Persson 2015). The heterogeneous environment featuring 15 habitats per the Habitats Directive (Kindvall et al. 2022), has seen intensified livestock grazing since 2000 (Kindvall et al. 2022). Old forests, primarily pine, are scattered throughout the region. The western and northeastern borders are marked by agricultural fields commonly treated with pesticides and fertilisers.

Fig. 1
figure 1

A red dot denotes the location of the study area—the spatial distribution of hectare grids with captured target species on Gotland, southeastern Sweden. Panel A depicts the geographic dispersion of Marsh Fritillary (Euphydryas aurinia), Panel B of Apollo (Parnassius apollo), and Panel C of Large Blue (Phengaris arion). In each panel, the top sub-panel illustrates the number of departures per grid, while the bottom sub-panel delineates the number of arrivals per hectare grid. The colour gradient, transitioning from blue to purple, quantitatively represents the range of recorded departures and arrivals, with purple indicating higher numbers. Photographs Markus Franzén

Full size image

Description of studied species

We examined three butterfly species: E. aurinia, P. apollo, and P. arion. These species face rapid declines globally and significant conservation challenges at regional and local levels in Europe. They are protected within the European Union (EU) under the Habitats Directive (Council Directive 92/43/EEC). Parnassius apollo is near threatened in Europe, P. arion is endangered, and E. aurinia is red-listed in several European countries (Maes et al. 2019). Notably, they can be locally abundant in areas of Gotland, enabling in-depth population and movement analysis (Franzen et al. 2022; Franzén et al. 2022, 2024; Johansson et al. 2020; Sunde et al. 2024).

Phengaris arion, a blue butterfly spanning 32–42 mm (Fig. 1), is native to the western Palaearctic. Thriving in dry calcareous grasslands and Alvar in our study region, this univoltine butterfly is active from July to August. The larvae, feeding on Thymus serpyllum, are adopted by Myrmica ants, feeding parasitically on ant broods (Thomas and Wardlaw 1992; Eliasson et al. 2005).

Parnassius apollo, a white butterfly in the Papilionidae family with a 73–87 mm wingspan (Fig. 1), populates areas similar to P. arion. Preferring areas with bare rocks, it thrives on open alvar terrain in our study area. P. apollo has experienced a notable decline since the 1950s, primarily due to habitat loss and environmental changes. Its adult flight period occurs from June to August and visits flowering Thymus serpyllum and Centaurea scabiosa, while the larvae feed on Sedum album from April to June (Eliasson et al. 2005).

Euphydryas aurinia, an orange-to-brown butterfly with a 33–48 mm wingspan (Fig. 1), ranges from northern Africa to China. It occurs on fens and ungrazed grasslands in our study area and is active from May to June. The larvae feed on Succisa pratensis, creating silk shelters and hibernating in vegetation, resuming feeding in spring (Eliasson et al. 2005).

Collection of butterfly occupancy and movement data

Our study utilised a systematic grid-based approach, covering the entire designated area. We overlaid a fishnet grid, each square measuring 100 × 100 m2, across the study area, resulting in 3430-hectare grids subjected to butterfly surveys. The surveys spanned different years for each species for logistical reasons: E. aurinia in 2017 across 1330 grids, P. apollo in 2019 across 2359 grids, and P. arion in 2020 over 2256 grids. The timing of our fieldwork was meticulously planned to coincide with the flight periods of the respective species. The survey for E. aurinia lasted 33 days in 2017, P. apollo spanned 52 days in 2019, and P. arion lasted 26 days in 2020. Surveys were conducted daily between 8 a.m. and 6 p.m. during suitable weather by up to ten trained personnel daily. Upon each butterfly’s capture, it was individually marked and identified by species and sex. The capture location and timestamp were directly and automatically recorded into our database through the Field Maps application (ESRI). We avoided surveys during rain or temperatures below 14 °C. Each hectare grid was systematically checked daily or every other day, and we employed a rotation strategy to ensure varied personnel surveyed grids at different times. In total, 14, 18, and nine field workers assisted with the E. aurinia, P. apollo, and P. arion surveys.

Land cover data extraction

Utilising the Swedish land cover database (Anonymous 2020), we procured data on three crucial land cover variables: forest cover, vegetated open land coverage, and the ground moisture index. For each hectare grid, we calculated the proportion of forest cover and vegetated open land and the average ground moisture index to be employed in subsequent modelling.

Assessing density-dependent movements: proportions of departures (emigration) and arrivals (immigration) per hectare grid

We defined two primary metrics per hectare grid to evaluate density dependence in butterfly movement: ‘arrivals’ and ‘departures’. The departure fraction for grid “1” was calculated by dividing the number of butterflies leaving grid “1” by the overall recaptures in grid “1”. Similarly, the arrival fraction for grid “1” was calculated by dividing the number of recaptures of butterflies not originally marked in grid “1” but found there later (immigration) by the overall recaptures in grid “1”.

Statistical analysis

Analysis of land cover and density-dependent movements

To assess potential land cover and density-dependent butterfly movements within hectare grids, we utilised Generalised Linear Models (GLMs) implemented through the glm function in the stats package, part of R’s core distribution. These models examined the relationship between the proportions of departures and arrivals (examined separately) and the four continuous predictors: forest cover vegetated open land coverage, ground moisture index, and population density (quantified as the number of individuals per visit) within the hectare grid. Two GLMs were conducted for each species—one for departures and one for arrivals. Models were manually selected based on the Akaike Information Criterion (AIC), with the final model boasting the lowest AIC presented. Given overdispersion and the proportions of departures (emigration) and arrivals (immigration) as dependent variables, a quasibinomial distribution was applied. This distribution corrects variance, making more precise statistical inferences. (Dunn et al. 2018). Visualisation tools from the ggplot2 and ggeffects packages facilitated model interpretation by presenting predicted probabilities for different population densities and capturing events and species (Lüdecke 2018). All statistical analyses were performed using R version 4.3.1 (R Core Team 2023).

To specifically investigate the dynamics of arrivals and departures associated with longer-distance movements, we refined our dataset to include only those events that surpassed species-specific movement thresholds. These thresholds were established based on the average movement distances for Euphydryas aurinia (942 m), Parnassius apollo (735 m), and Phengaris arion (594 m), respectively. This filtration allowed us to isolate and examine the subset of movements that constitute significant (longer-term) dispersal events rather than shorter routine, localised foraging, sun basking or mating activities. Subsequent analyses utilised Generalized Linear Models (GLMs) as described above to assess the influence of grid characteristics and population densities on these arrivals and departures that resulted from long-distance movement events.

Results

Departure and arrival patterns

The uniquely marked individuals per species included 10,161 E. aurinia in 2017, 5902 P. apollo in 2019, and 160 P. arion in 2020. Table 1 presents the density of butterflies (average individuals marked per visit per grid), recapture rates, arrivals, and departures.

Table 1 The average density of individuals marked per visit per grid, recapture rates (with total individuals recaptured), and the number of arrivals and departures documented for each species
Full size table

Movements in relation to land cover

For P. apollo, we identified a negative correlation between departures from the hectare grid and forest cover and a positive correlation with the ground moisture index (Fig. 2; Table 2). Similarly, the pattern for arrivals of P. apollo mirrored the findings for departures to forest cover and ground moisture index (Fig. 2; Table 2). For E. aurinia, an increase in the proportion of vegetated ‘other open land’ was correlated with an increased frequency of departures from the grids and a similarly positive relationship with the ground moisture index (Fig. 2; Table 2). These correlations were reciprocated in the frequency of arrivals, with significant values for vegetated ‘other open land’ and ground moisture index (Fig. 2; Table 2). In P. arion, neither arrivals nor departures from the grids were significantly associated with any of the three habitat features that we investigated (Fig. 2; Table 2). None of the examined factors statistically significantly influenced longer-distance arrivals or departures among the studied butterfly species.

Fig. 2
figure 2

Model coefficients representing the estimates of different predictors ± SE for the three butterfly species. Each point denotes the estimate for a predictor, with the colour indicating the specific butterfly species: E. aurinia (red), P. apollo (black), and P. arion (blue). The associated error bars signify the standard error of these estimates. Predictors are displayed on the y-axis, while the x-axis represents the estimated values. Circles = departures, and triangles = arrivals

Full size image
Table 2 Results from the final best generalised linear model examining the influence of land cover and population density (number of butterflies per visit) on the proportion of departures and arrivals (in hectare grids)
Full size table

Density-dependent movements

Both E. aurinia and P. apollo showed strong positive trends for arrivals and departures in relation to population density, as displayed in Table 2. However, P. arion also leaned towards similar trends, although it did not reach traditional levels of statistical significance. Despite this, including these variables improved the model’s overall fit, according to the AIC. Arrivals and departures filtered to only include longer-distance movements were not significantly related to population density.

Discussion

Departure and arrival patterns of two of the three butterfly species studied were intricately tied to land cover characteristics and local population densities. The dichotomy between within-site and between-site movements suggests distinct ecological and behavioural drivers at play. Within-site movements, predominantly within the same hectare grid (< 100 m), are closely tied to daily survival strategies, such as temperature regulation, foraging for nectar, and seeking host plants, reflecting localised resource utilisation (Warren et al. 2001; Dover and Settele 2009). In contrast, between-site movements here defined as moving from and to another hectare grid, although less frequent, suggest a response to land cover properties, possibly driven by the need for suitable breeding sites, avoidance of predation, or adaptation to environmental changes (Schtickzelle et al. 2006). Although our study did not find significant determinants for long-distance movements, this aspect remains crucial for understanding metapopulation dynamics and gene flow. The lack of significant results may be attributed to the relative rarity of long-distance dispersal events, which are often stochastic and influenced by factors beyond the scope of our study design (Kindvall 1999; Hanski 1994, 1998; Hanski et al. 2011). Landscape barriers, weather conditions, and individual variation in dispersal capacity could all contribute to the complexity of long-distance movements (Alerstam et al. 2003). Future research following each movement of individual butterflies may be necessary to elucidate patterns in long-distance dispersal for these species.

When considering these species, conservation strategies must account for the nuanced interplay between habitat attributes and local butterfly abundances. For P. apollo, the evident disinclination towards tree-covered areas aligns well with its established preference for open terrains replete with bare rocks (Nakonieczny et al. 2007; Sunde et al. 2024). The species has been noted for its selective habitat choices, primarily seeking areas with minimal vegetation, a behavioural trait potentially driven by evolutionary and ecological pressures (Fred et al. 2006). Euphydryas aurinia, on the other hand, displayed a strong inclination towards vegetated open lands, supporting findings from previous studies which suggest this species thrives in such habitats due to the availability of specific resources or potential mating opportunities (Warren et al. 1994; Wahlberg et al. 2002). The ground moisture index also played a pivotal role, which is unsurprising given the butterfly’s predilection for rich fens and ungrazed grasslands, habitats often associated with higher moisture levels (Botham et al. 2011). Lastly, the P. arion exhibited patterns subtly influenced by population densities rather than explicit land cover types. This could imply a more complex behavioural framework, possibly driven by interspecific interactions, competition, or other ecological dynamics not directly evaluated in this study (Hayes 2015; Vilbas et al. 2015).

Parnassius apollo and E. aurinia arrived and departed from grids with certain land cover properties. These movements in butterfly species are not merely a linear trajectory from unsuitable to suitable habitats; their activities seem to be influenced by a complex matrix of factors. The decision to depart from or arrive in a particular area could be driven by the heterogeneity of microhabitats within a larger ‘suitable’ habitat. For instance, butterflies may continually seek optimal microsites within a favourable landscape to maximise access to resources such as nectar, mates, or oviposition sites (Hanski 1994; Schultz 2001). Moreover, these butterflies' movements might be intricately tied to their behavioural ecology. Territoriality, mate-seeking behaviours, and avoiding intraspecific competition could all play pivotal roles in their frequent movements (Baguette et al. 1996; Brunzel 2002). It is also worth considering that some areas, while rich in resources, might also be hotspots for predators or parasitoids, prompting butterflies to move frequently to balance resource acquisition with predation risk (Kruess and Tscharntke 2000; Eliasson and Shaw 2003). Furthermore, butterflies' cognitive abilities should be considered. They may engage in exploratory flights within suitable habitats to gather information and reevaluate the quality of their surroundings (Dukas 1998; Couto et al. 2023). In doing so, they continually optimise their choices, even within a habitat broadly categorised as ‘suitable’. Lastly, environmental cues, including temperature, humidity, and light fluctuations, could prompt temporary movements even within favourable habitats. Such transient departures and arrivals might be strategies to optimise thermoregulation or avoid transient adverse conditions (Porter 1982, 1984; Kuussaari et al. 1996; Braem and Van Dyck 2021; Franzen et al. 2022).

Forested areas are characterised by dense tree canopies, which can significantly reduce sunlight penetration to the forest floor. The resultant cool, shaded microclimate is typically less hospitable to many butterfly species, particularly those adapted to open habitats (Dennis et al. 2004). The diminished sunlight also translates to fewer flowering plants, which are primary nectar sources for butterflies (Clausen et al. 2001). Consequently, the negative association we observed between forest cover and butterfly arrivals and departures aligns with the understanding that dense forests are less preferred if not outright avoided, due to their suboptimal conditions (Franzén and Nilsson 2008; Sunde et al. 2024). Further, in stark contrast to forests, open vegetated areas offer abundant sunlight, facilitating the growth of many flowering plants. Such landscapes become veritable oases for butterflies, providing them with a rich supply of nectar—a crucial energy source (Tscharntke et al. 2002). Our findings of a positive relationship between open vegetated land and butterfly movements echo this ecological reality. The open land offers sustenance and promotes mate-finding and oviposition activities due to increased butterfly visibility (Thomas et al. 1996; Konvicka and Kuras 1999).

Moist conditions are generally favourable for butterflies for several reasons. First, damp environments can support a diverse range of plant species, enhancing habitat quality (Maclean et al. 2015). Moreover, moist areas can also serve as ‘mud-puddling’ sites, where butterflies extract essential minerals from damp soils—a behaviour observed especially in males (Molleman 2010). Our findings suggest a complex relationship between ground moisture and butterfly movement that varies among species. While some butterflies prefer arid habitats, ground moisture may serve as a proxy for other crucial variables like floral abundance or microhabitat structures (Cabral et al. 2017; Franzen et al. 2022). Our results indicate that moisture conditions indirectly influence habitat selection and movement patterns of E. aurinia and P. apollo, whose larval food plants and ant nests, respectively, are affected by soil moisture (Thomas et al. 1989; Fred et al. 2006; Franzén et al. 2022; Kindvall et al. 2022).

As a mobile species and avid flower visitor (Franzén et al. 2024), P. apollo's departures and arrivals may be driven by the pursuit of abundant floral resources or favorable microclimatic conditions for thermoregulation in areas of higher ground moisture (Fred et al. 2010; Franzen et al. 2022). Intriguingly, P. arion diverged from the behavioural trends exhibited by the other butterfly species under consideration. While variables such as forest cover, open vegetated land, and ground moisture index were less impactful, population density emerged as a salient predictor for the movements of P. arion. This divergence could potentially be attributed to the smaller sample size and fewer individual specimens studied, thereby reducing statistical power to detect nuanced habitat preferences (Thomas et al. 2009). Additionally, the populations of these species are recovering from a recent drought, which could be a confounding variable influencing movement patterns. It is plausible that P. arion's ecological requirements (depends on certain nectar plants and host ants) or behaviours predispose it to be more sensitive to conspecific density over broader habitat characteristics. The ecological specificity of P. arion may render it more responsive to variations in conspecific density rather than to general habitat features. This interpretation aligns with the findings of Thomas et al. (2011), who observed that P. arion exhibits a heightened sensitivity to conspecific presence, potentially influencing its dispersal and habitat utilisation patterns more than the physical attributes of the environment. Moreover, Osváth-Ferencz et al. (2017) state that the unique life-cycle dependencies and social interactions within P. arion populations accentuate the role of density dynamics over environmental heterogeneity in determining their spatial distribution. Given the limited dataset for this species, we advocate for further extensive studies to elucidate the specific habitat preferences of P. arion.

Our study unearthed a consistent trend of positive density-dependent mobility across all three butterfly species, resonating with patterns identified in various flora and fauna studies (Rodrigues and Johnstone 2014). The annals of ecological research have documented both positive (Nowicki and Vrabec 2011) and negative density-dependent mobilities (Gilbert and Singer 1973; Brown and Ehrlich 1980; Ims and Andreassen 2005; Støen et al. 2006; Konvicka et al. 2012). Intriguingly, the same species, such as the E. aurinia and Glanville fritillary butterfly, Melitaea cinxia, have demonstrated contrasting mobility depending on circumstances (Kuussaari et al. 1996; Enfjäll and Leimar 2005; Johansson et al. 2022). A particularly captivating observation from our study is the propensity of these butterflies to leave high-density regions, only to later settle in areas of similar density. This challenges the conventional ecological wisdom, which posits that species migrate from areas of high to low density to mitigate competition for resources (Solomon 1949). Instead, our data suggests that these butterflies might be navigating between densely populated grids in pursuit of prime habitats (Dunning et al. 1992), evading predators (Lima and Dill 1990), or responding to intrinsic territorial and behavioural impulses (Bowler and Benton 2005). Future endeavours in this domain should delve deeper into individual butterfly movement patterns and map resource availability (Dennis et al. 2003), and explore potential predation threats.

Conclusion and future directions

In synthesising these findings, it becomes clear that while overarching patterns can be discerned, each species exhibits unique responses to the environment. Such distinctions underscore the importance of tailored conservation strategies. When working towards conserving these endangered species, it is imperative to consider both the broader landscape changes and the intricate species-specific behaviours in response to their immediate surroundings (Jones & Harrison, 2015; Smith et al., 2010).