Ecologists are challenged to construct models of the biological consequences of habitat loss and fragmentation. Here, we use a metapopulation model to predict the distribution of the Glanville fritillary butterfly during 22 years across a large heterogeneous landscape with 4,415 small dry meadows. The majority (74%) of the 125 networks into which the meadows were clustered are below the extinction threshold for long-term persistence. Among the 33 networks above the threshold, spatial configuration and habitat quality rather than the pooled habitat area predict metapopulation size and persistence, but additionally allelic variation in a SNP in the gene Phosphoglucose isomerase (Pgi) explains 30% of variation in metapopulation size. The Pgi genotypes are associated with dispersal rate and hence with colonizations and extinctions. Associations between Pgi genotypes, population turnover and metapopulation size reflect eco-evolutionary dynamics, which may be a common feature in species inhabiting patch networks with unstable local dynamics.

Habitat fragmentation and climate change are both prominent manifestations of global change, but there is little knowledge on the specific mechanisms of how climate change may modify the effects of habitat fragmentation, for example, by altering dynamics of spatially structured populations. The long‐term viability of metapopulations is dependent on independent dynamics of local populations, because it mitigates fluctuations in the size of the metapopulation as a whole. Metapopulation viability will be compromised if climate change increases spatial synchrony in weather conditions associated with population growth rates. We studied a recently reported increase in metapopulation synchrony of the Glanville fritillary butterfly (Melitaea cinxia) in the Finnish archipelago, to see if it could be explained by an increase in synchrony of weather conditions. For this, we used 23 years of butterfly survey data together with monthly weather records for the same period. We first examined the associations between population growth rates within different regions of the metapopulation and weather conditions during different life‐history stages of the butterfly. We then examined the association between the trends in the synchrony of the weather conditions and the synchrony of the butterfly metapopulation dynamics. We found that precipitation from spring to late summer are associated with the M. cinxia per capita growth rate, with early summer conditions being most important. We further found that the increase in metapopulation synchrony is paralleled by an increase in the synchrony of weather conditions. Alternative explanations for spatial synchrony, such as increased dispersal or trophic interactions with a specialist parasitoid, did not show paralleled trends and are not supported. The climate driven increase in M. cinxia metapopulation synchrony suggests that climate change can increase extinction risk of spatially structured populations living in fragmented landscapes by altering their dynamics.

Dispersal is important for determining both species ecological processes, such as population viability, and its evolutionary processes, like gene flow and local adaptation. Yet obtaining accurate estimates in the wild through direct observation can be challenging or even impossible, particularly over large spatial and temporal scales. Genotyping many individuals from wild populations can provide detailed inferences about dispersal. We therefore utilized genomewide marker data to estimate dispersal in the classic metapopulation of the Glanville fritillary butterfly (Melitaea cinxia L.), in the Åland Islands in SW Finland. This is an ideal system to test the effectiveness of this approach due to the wealth of information already available covering dispersal across small spatial and temporal scales, but lack of information at larger spatial and temporal scales. We sampled three larvae per larval family group from 3732 groups over a six‐year period and genotyped for 272 SNPs across the genome. We used this empirical data set to reconstruct cases where full‐sibs were detected in different local populations to infer female effective dispersal distance, that is, dispersal events directly contributing to gene flow. On average this was one kilometre, closely matching previous dispersal estimates made using direct observation. To evaluate our power to detect full‐sib families, we performed forward simulations using an individual‐based model constructed and parameterized for the Glanville fritillary metapopulation. Using these simulations, 100% of predicted full‐sibs were correct and over 98% of all true full‐sib pairs were detected. We therefore demonstrate that even in a highly dynamic system with a relatively small number of markers, we can accurately reconstruct full‐sib families and for the first time make inferences on female effective dispersal. This highlights the utility of this approach in systems where it has previously been impossible to obtain accurate estimates of dispersal over both ecological and evolutionary scales.

Species distribution models are the tool of choice for large-scale population monitoring, environmental association studies and predictions of range shifts under future environmental conditions. Available data and familiarity of the tools rather than the underlying population dynamics often dictate the choice of specific methodespecially for the case of presence-absence data. Yet, for predictive purposes, the relationship between occupancy and abundance embodied in the models should reflect the actual population dynamics of the modelled species. To understand the relationship of occupancy and abundance in a heterogeneous landscape at the scale of local populations, we built a spatio-temporal regression model of populations of the Glanville fritillary butterfly Melitaea cinxia in a Baltic Sea archipelago. Our data comprised nineteen years of habitat surveys and snapshot data of land use in the region. We used variance partitioning to quantify relative contributions of land use, habitat quality and metapopulation covariates. The model revealed a consistent and positive, but noisy relationship between average occupancy and mean abundance in local populations. Patterns of abundance were highly variable across years, with large uncorrelated random variation and strong local population stochasticity. In contrast, the spatio-temporal random effect, habitat quality, population connectivity and patch size explained variation in occupancy, vindicating metapopulation theory as the basis for modelling occupancy patterns in fragmented landscapes. Previous abundance was an important predictor in the occupancy model, which points to a spillover of abundance into occupancy dynamics. While occupancy models can successfully model large-scale population structure and average occupancy, extinction probability estimates for local populations derived from occupancy-only models are overconfident, as extinction risk is dependent on actual, not average, abundance. Long-term demographic surveys reveal a consistent relationship between average occupancy and abundance within local populations of a butterfly metapopulation Torsti Schulz, Jarno Vanhatalo and Marjo Saastamoinen T. Schulz (https://orcid.org/0000-0001-8940-9483) ✉ (torsti.schulz@helsinki.fi), J. Vanhatalo (https://orcid.org/0000-0002-6831-0211) and M. Saastamoinen (https://orcid.org/0000-0001-7009-2527),

Metapopulation dynamics – patch occupancy, colonization and extinction – are the result of complex processes at both local (e.g. environmental conditions) and regional (e.g. spatial arrangement of habitat patches) scales. A large body of work has focused on habitat patch area and connectivity (area‐isolation paradigm). However, these approaches often do not incorporate local environmental conditions or fully address how the spatial arrangement of habitat patches (and resulting connectivity) can influence metapopulation dynamics. Here, we utilize long‐term data on a classic metapopulation system – the Glanville fritillary butterfly occupying a set of dry meadows and pastures in the Åland islands – to investigate the relative roles of local environmental conditions, geographic space and connectivity in capturing patch occupancy, colonization and extinction. We defined connectivity using traditional measures as well as graph‐theoretic measures of centrality. Using boosted regression tree models, we find roughly comparable model performance among models trained on environmental conditions, geographic space or patch centrality. In models containing all of the covariates, we find strong and consistent evidence for the roles of resource abundance, longitude and centrality (i.e. connectivity) in predicting habitat patch occupancy and colonization, while patch centrality (connectivity) was relatively unimportant for predicting extinction. Relative variable importance did not change when geographic coordinates were not considered and models underwent spatially stratified cross‐validation. Together, this suggests that the combination of regional‐scale connectivity measures and local‐scale environmental conditions is important for predicting metapopulation dynamics and that a stronger integration of ideas from network theory may provide insight into metapopulation processes.