The effect of conspecific attraction on metapopulation dynamics

Random dispersal direction is assumed in all current metapopulation models. This assumption is called into question by recent experiments demonstrating that some species disperse preferentially to sites occupied by conspecifies. We incorporate conspecific attraction into two metapopulation models which differ in type of dispersal, the Levins model and a two-dimensional stepping-stone model. In both models, conspecific attraction lowers the proportion of occupied habitat patches within a metapopulation at equilibrium.     

Distribution dynamics of a great bustard metapopulation throughout a decade: influence of conspecific attraction and recruitment

Dispersing individuals can use conspecifics as indicators of habitat quality and aggregate at traditionally occupied sites, leaving other favourable patches unoccupied. Here we test the predictions of the conspecific-based habitat selection hypothesis on a Spanish great bustard (Otis tarda) metapopulation, currently fragmented due to recent human-induced habitat changes. The number of birds had increased by 23% between 1988 and 1998, but not consistently among leks. Leks that were large in 1988 increased, while those that were small decreased, which suggests that dispersing individuals used the numbers of conspecifics as cues for breeding-site selection. Moreover, leks with high productivity increased, while those with low productivity decreased. Finally, lek distribution was markedly stable throughout the decade, with no establishment of new leks, and suitable habitat patches remained unoccupied, as predicted by the conspecific attraction hypothesis. These results were corroborated by a simulation model which incorporated natal dispersal rates between leks as obtained through radio-tracking of 15 birds that survived throughout their 4-year dispersal period. In conclusion, in spite of the apparent increase in total numbers throughout the decade, both conspecific attraction and local differences in reproductive success contributed to a more aggregated distribution, increasing the species' vulnerability to local catastrophes, and the risks of reduced genetic diversity and extinction of small leks.     

Testing the conspecific attraction hypothesis with Dupont's Larks,a resident species of songbird in central Spain

Some birds use social cues, such as the presence of conspecifics, when selecting breeding habitat. This phenomenon, known as conspecific attraction, has been well‐documented in migratory species, but has not been assessed for resident species of birds. We used Dupont's Larks (Chersophilus duponti) as a model species to determine if conspecific attraction plays a role in habitat selection by resident species of birds. At our study site in Soria province in central Spain, we monitored two potential habitat patches and one managed site where management actions had provided apparently suitable habitat. At each site, we broadcast recordings of the songs and calls of male Dupont's Larks, and monitored their presence during the breeding season and dispersal period in 2018 using automated recorders and field surveys. No birds were attracted to our study sites. Our results suggest that management of patches of suitable habitat should occur close to areas (within 1 km) already occupied by Dupont's Larks to encourage natural colonization because, based on our results, playback of conspecific vocalizations may not attract the species to new breeding areas. However, additional studies are needed before drawing conclusions about the effectiveness of conspecific attraction for this and other resident species of birds.     

Patterns of commonness and rarity in central European birds: reliability of the core-satellite hypothesis within a large scale

The frequency distribution of species’ area of occupancy is often bimodal, most species being either very rare or very common in terms of number of occupied sites. This pattern has been attributed to the nonlinearity associated with metapopulation dynamics of the species, but there are also other explanations comprising sampling artifact and frequency distribution of suitable habitats. We tested whether the bimodal frequency distribution of occupied squares in central European birds could be derived solely from the frequency distribution of species population sizes (i.e. the sampling artifact hypothesis) or from the spatial distribution of their preferred habitats. Both models predict high proportion of very common species, i.e. the right side of frequency distribution. Bimodality itself is well predicted by models based on random placement of individuals according to their abundances but neither model predicts the observed prevalence of rare species. Even the combined models that assume random placement of individuals within the squares with suitable habitat do not predict such a high proportion of rare species. The observed distribution is more aggregated, rare species occupying a smaller portion of suitable habitat than predicted on the basis of their abundance. The pattern is consistent with metapopulation processes involving local population extinctions. The involvement of these processes is supported by two further observations. First, species rarity is associated with significant population trend and/or location on the edge of their ranges within central Europe, both situations presumably associated with metapopulation processes. Second, suitable habitats seem to be either saturated or almost unoccupied, which is consistent with the predictions of the metapopulation model based on nonlinear dynamics of extinction and colonization. Although the habitat suitability is an important determinant of species distribution, the rarity of many species of birds within this scale of observation seems to be affected by other factors, including local population extinctions associated with fragmentation of species’ habitats.     

Conspecific attraction in a grassland bird, the Baird's Sparrow

ABSTRACT.   Territorial songbirds generally use song to defend territories and attract mates, but conspecific song may also serve as a cue to attract other male songbirds to a breeding site. Although known to occur in some colonial and forest-associated species, only recently have investigators examined conspecific attraction in grassland species. We used a playback experiment to examine the possible role of conspecific attraction for males searching for potentially suitable breeding habitat in a grassland specialist, the Baird's Sparrow ( Ammodramus bairdii ). Experimental playback plots and control plots with similar landscape and vegetation characteristics were established at two sites in North Dakota. Baird's Sparrows colonized three of six experimental plots and none of six control plots. Males on experimental plots established territories adjacent to the playback stations and were sometimes observed counter-singing with the playback of conspecific songs. Vegetation characteristics were similar on all study plots, and did not explain differences in bird density on our treatment plots. Although we found that playback of conspecific songs attracted male Baird's Sparrows to previously unoccupied, potentially suitable habitat, further experiments are needed to examine the importance of conspecific attraction relative to other cues that birds may use, such as vegetation features. The conservation and management implications of conspecific attraction are not completely understood, but the presence of conspecifics should be considered as a potential cue in habitat selection by all species of birds.     

Conserving extirpated sites: using habitat quality to manage unoccupied patches for metapopulation persistence

Conservation of metapopulations requires managing extirpated sites, particularly with current threats of increased fragmentation and displacement from global warming. Determining the habitat requirements of threatened species and how they relate to defining characteristics of occupied and unoccupied sites is key to managing suitable habitat in extirpated patches. Due to habitat destruction and degradation, the endangered Ohlone tiger beetle (Cicindela ohlone) is found in only five sites of a once more extensive metapopulation in Santa Cruz County, California. To determine the role of habitat quality in classifying sites, I measured vegetation and ground cover as well as plant and soil composition in sites in which C. ohlone are present, extirpated, and absent. I used conditional inference trees to determine what habitat factors significantly predicted the different sites types. I also analyzed habitat characteristics within present sites to determine factors that predicted egg-laying habitat. As isolation has been shown to be an important driver of metapopulation patch extirpation, I tested the spatial autocorrelation of C. ohlone occupancy to determine if extirpated patches were significantly isolated. Habitat characteristics successfully differentiated nearly 90 % of extirpated plots, which were not isolated from occupied sites. Sites in which C. ohlone are currently present were classified as having at least 10 % cover of bare ground, high forb cover, low litter cover and depth, and high soil bulk density, characteristics that extirpated sites lacked. I illustrate how the defining characteristics could be used to manage habitat in extirpated and absents sites for potential recolonization or translocation, which is vital for metapopulation persistence.     

Hypericum cumulicola demography in unoccupied and occupied Florida scrub patches with different time‐since‐fire

1 Metapopulation models predict that unoccupied, but suitable, patches will exist for species subject to extinction and colonization dynamics. We compared the demographic responses of Hypericum cumulicola , a rare herbaceous species almost entirely restricted to Florida rosemary scrub, when transplanted to occupied or unoccupied patches.
2 Seedlings were transplanted and seeds buried into Florida rosemary scrub patches differing in time since last fire, and in the presence or absence of H. cumulicola. We used a replicated, factorial design to place the transplants and seeds in the field, and monitored their performance for 18 months.
3 Neither time‐since‐fire nor prior H. cumulicola site occupancy affected survival of transplants. Only time‐since‐fire affected growth. Time‐since‐fire, H. cumulicola occupancy, and their interaction affected reproductive effort, but these effects were not consistent between years.
4 Flowering and seed production led to subsequent seedling recruitment near transplants, mainly in recently burned sites. Genetic screening of transplants and seedlings showed that transplants in occupied sites could have crossed with nearby resident plants, but that offspring in sites previously unoccupied were likely to have been parented only by nearby transplants.
5 Seeds buried, and later exhumed, germinated after 1 or 2 years of burial, demonstrating a persistent soil seed bank from which populations could recover after fire. Neither time‐since‐fire nor H. cumulicola occupancy affected seed dormancy or germination.
6 Similar demography in unoccupied and occupied patches suggests that the patchy pattern of site occupancy by H. cumulicola is probably due to limited dispersal and periodic extinction, especially associated with long fire‐free intervals. Conservation measures need to protect unoccupied patches to allow metapopulation dynamics and persistence.     

Spatial population structure in the banner-tailed kangaroo rat,Dipodomys spectabilis

I attempted to characterize spatial units of local dynamics and dispersal in banner-tailed kangaroo rats (Dipodomys spectabilis), to determine if spatial structure influenced population dynamics in the way predicted by current metapopulation models. D. spectabilis exhibited a hierarchical spatial structure. Local populations that appeared as discrete entities on a scale of kilometers were subdivided into clusters of mounds on a scale of meters. This structure, however, cannot be characyerized in terms of the discrete habitat patches envisioned by the metapopulation models. Occupied areas were statistically distinguishable from the surrounding matrix, but this difference was only quantitative. There were no discrete boundaries between occupied areas and the matrix. Habitat within occupied areas was heterogeneous, and occupied areas in different locations were statistically distinguishable from each other. High heterogeneity within occupied areas, and high contrast among them, make it difficult to define what is a suitable habitat patch for D. spectabilis. On a smaller spatial scale, there was significant aggregation of resident mounds within occupied areas. These aggregations, however, do not correspond to discrete habitat patches. Rather, they appear to result from an interaction between fine-scale habitat heterogeneity and limited dispersal due to natal philopatry and low adult vagility. These complications make it difficult to identify habitat patches independent of the species' distribution. For species like D. spectabilis that are patchily distributed but do not occupy discrete habitat patches, a patch occupancy approach does not seem appropriate for describing spatial structure. Hierarchical spatial structure underscores the need for a framework that incorporates multiple scales of spatial structure, rather than one that pre-imposes a single spatial scale as being important for population dynamics. A framework that (i) considers patchiness as a combination of both habitat heterogeneity, and life-history and behavioral characteristics, and (ii) incorporates hierarchical spatial structure, appears to be the most suitable for conceptualizing spatial dynamics of behaviorally complex vertebrates such as D. spectabilis.     

Identifying Core Areas in a Species’ Range Using Temporal Suitability Analysis: an Example Using Little Bustards <Emphasis Type="Italic">Tetrax Tetrax</Emphasis> L. in Spain

Variations in habitat quality impact on breeding success, leading to strong selection pressure for the best sites to be occupied first during a population increase and last during a decline. Coupled with dispersal and metapopulation processes, the result is that snapshot surveys of wildlife distributions may fail to reveal core areas that conservation seeks to protect. At a local scale, territory occupancy is a good indicator of quality but data are not readily available to assess occupancy for rarer species, in remote areas, and over large spatial extents. We introduce temporal suitability analysis as a way to generate an analogue of occupancy from a single survey and illustrate it using data on the little bustard in Spain. We first used Generalised Additive Modelling (GAM) to build a predictive distribution model using Geographic Information System (GIS) coverages and satellite imagery, and then applied the model retrospectively to a time series of satellite images to produce one distribution map for each year. These annual maps differed in the extent of Spain predicted as suitable for little bustards. By overlaying the maps, we identified areas predicted as suitable in one to n years. We show that this temporal suitability map correlates with a conventional habitat suitability map based on a single year but contains extra information on hierarchical use of habitats and the lag between suitability and use. The technique may be applied at a variety of spatial scales to reveal changes in expected occupancy as land use or external factors determining land cover types vary over time.     

Propagule vs. niche limitation: untangling the mechanisms behind plant species' distributions

Distinguishing the roles of propagule limitation and niche requirements in controlling plant species distributions is important for understanding community structure, invasion, and restoration. We used species distribution models based on plant and environmental survey data to assess the strength of species' affinities for particular environmental conditions. We hypothesized that species with statistically detectable environmental requirements were primarily niche-limited, while species with weak habitat affinities were primarily propagule-limited. We tested this hypothesis via a seeding experiment in which we compared species' reproductive fitness in occupied and unoccupied sites. Species that appeared to be niche-limited based on distribution models had lower fitness when planted in unoccupied sites, while species that models suggested were propagule-limited had equivalent fitness when planted in occupied and unoccupied sites. Our results demonstrate that within a single community, both species limited primarily by niche availability or primarily by propagule availability can be identified using observational data.     

Identification of suitable unoccupied habitats in metapopulation studies using co-occurrence of species

This study proposed a new quantitative technique to identify suitable but unoccupied habitats for metapopulation studies in plants. It is based on species composition at the habitat and knowledge of species co-occurrence patterns. It uses data from a large phytosociological database as a background for estimating species co-occurrence patterns. If such a database is not available, the technique can still be applied using the same data for which the prediction is done to estimate the species co-occurrence pattern. Using the technique we were able to indicate suitable unoccupied habitats and differentiate them from the unoccupied unsuitable ones. We also identified habitats with low probability of being suitable that were occupied. Compared to a direct approach of identification of suitable habitats, which involves introduction of a species to the habitat and studying its performance, the approach presented here is much easier to apply and can provide extensive information on habitat suitability for a range of species with much less effort and time needed.     

Winter habitat use by Loggerhead Shrikes (Lanius Ludovicianus) in Mexico: separating migrants from residents using stable isotopes

Several species of migrant birds overlap in range on their wintering grounds with non-migrant conspecifics or other species that occupy a similar niche. Very little is known whether such overlap results in competition and subsequent habitat segregation since it is usually impossible to separate resident from migrant individuals. The Loggerhead Shrike (Lanius ludovicianus) is a declining grassland species in North America that winters in the southern United States and Mexico. Using stable-hydrogen isotope (δD) analysis of feathers, we identified resident and migrant shrikes wintering in northeastern Mexico based on a latitudinal gradient in precipitation and feather δD values. Indicator species analyses showed that migrants occupied areas where bare ground was less available than those occupied by residents, a pattern which held when a more restricted set of birds from the extremes of the δD distribution were considered. This provides evidence for conspecific habitat segregation. Habitat differences were also found between sites occupied by shrikes and apparently suitable but unoccupied sites. Shrikes occupied more open sites that contained shorter tall shrubs and huisache (Acacia farneasiana) and fewer tall shrubs, mesquite (Prosopsis glandulosa) and huisache than unoccupied sites. The availability of suitable winter habitat and the potential competition between migrants and residents may be factors that influence the population dynamics of migrant shrikes in North America.     

The influence of patch demographics on metapopulations, with particular reference to successional landscapes

The classical view of metapopulations relates the regional abundance of a species to the balance between the extinction and colonization dynamics of identical local populations. Species in successional landscapes may represent the most appropriate examples of classical metapopulations. However, Levins‐type metapopulation models do not explicitly separate population loss due to successional habitat change from other causes of extinction. A further complication is that the chance of population loss due to successional habitat change may be related to the age of a patch. I developed simple patch occupancy models to include succession and included consideration of patch age structure to address two related questions: what are the implications of changes in patch demographic rates and when is a move to a structured patch occupancy model justified? Age‐related variation in patch demography could increase or decrease the equilibrium fraction of the available habitat occupied by a species when compared to the predictions of an unstructured model. Metapopulation persistence was enhanced when the age class of patches with the highest species occupancy suffered relatively low losses to habitat succession. Conversely, when the age class of patches with the highest species occupancy also had relatively high successional loss rates, extinction thresholds were higher that would be predicted by a simple unstructured model. Hence age‐related variation in patch successional rate introduces biases into the predictions of simple unstructured models. Such biases can be detected from field surveys of the fraction of occupied and unoccupied patches in each age class. Where a bias is demonstrated, unstructured models will not be adequate for making predictions about the effects of changing parameters on metapopulation size. Thinking in successional terms emphasizes how landscapes might be managed to enhance or reduce the patch occupancy by any particular metapopulation     

Conspecific and heterospecific attraction in assessments of functional connectivity

Functional connectivity is known to have an important, positive influence on species persistence. Measurements of functional connectivity traditionally focus on structural attributes of landscapes such as the distance and matrix type between habitat patches as well as on how species interact with those structural attributes. However, we propose that the social behavior of a species, through conspecific and heterospecific attraction, will also impact connectivity by changing how dispersers move with respect to each other and occupied patches. We analyzed functional connectivity patterns using circuit and graph theory for golden-headed lion tamarins (Leontopithecus chrysomelas) in Brazil under three scenarios. In the first scenario, we looked at connectivity without the effects of attraction under varying maximum dispersal distance and ecological movement cost thresholds. In the second scenario, we allowed dispersers to travel over more hostile matrix than they normally would to reach an occupied patch. In the final scenario, we allowed dispersers to move only to occupied patches. We found that, according to the first scenario, range-wide functional landscape connectivity for golden-headed lion tamarins is low at realistic maximum dispersal distance and movement cost thresholds. Incorporating the effects of conspecific or heterospecific attraction would increase functional connectivity, in the case of scenario two, or decrease functional connectivity, in the case of scenario three. Because conspecific/heterospecific attraction can have an impact on movement for some species, this factor should be incorporated in assessments of functional connectivity patterns for social species and others where patch occupancy is likely to influence the movements of dispersers.     

Do invasive species undergo metapopulation dynamics? A case study of the invasive Caspian gull,Larus cachinnans,in Poland

Aim The mechanisms of initial dispersal and habitat occupancy by invasive alien species are fundamental ecological problems. Most tests of metapopulation theory are performed on local population systems that are stable or in decline. In the current study we were interested in the usefulness of metapopulation theory to study patch occupancy, local colonization, extinction and the abundance of the invasive Caspian gull (Larus cachinnans) in its initial invasion stages. Location Waterbodies in Poland. Methods Characteristics of the habitat patches (waterbodies, 35 in total) occupied by breeding pairs of Caspian gulls and an equal sample of randomly selected unoccupied patches were compared with t‐tests. Based on presence–absence data from 1989 to 2006 we analysed factors affecting the probability of local colonization, extinction and the size of local populations using generalized linear models. Results Occupied habitat patches were significantly larger and less isolated (from other habitat patches and other local populations) and were located closer to rivers than empty patches. The proximity of local food resources (fish ponds, refuse dumps) positively affected the occurrence of breeding pairs. The probability of colonization was positively affected by patch area, and negatively by distances to fish ponds, nearest habitat patch, nearest breeding colony and to a river, and by higher forest cover around the patch boundaries. The probability of extinction was lower in patches with a higher number of breeding pairs and with a greater area of islets. The extinction probability increased with distances to other local populations, other habitat patches, fish ponds and to refuse dumps and with a higher cover of forest around the patch boundaries. The size of the local population decreased with distances to the nearest habitat patch, local population, river, fish pond and refuse dump. Local abundance was also positively affected by the area of islets in the patch. Main conclusions During the initial stages of the invasion of Caspian gulls in Poland the species underwent metapopulation‐like dynamics with frequent extinctions from colonized habitat patches. The results prove that metapopulation theory may be a useful conceptual framework for predicting which habitats are more vulnerable to invasion.     

Temporal patch occupancy dynamics of the Siberian flying squirrel in a boreal forest landscape

Ability to predict species distribution in a landscape is of crucial importance for natural resource management and species conservation. Therefore, the understanding of species habitat requirements and spatio-temporal dynamics in occurrence is needed. We examined patch occupancy patterns of the Siberian flying squirrel Pteromys volans in northern Finland across a seven year study period. Forest patches dominated by mature spruce ( Picea abies ) in a study area (375 km²) were surveyed to monitor the presence or absence of the flying squirrel. The patch occupancy pattern was dynamic: about half of the habitat patches were occupied at least once during the study period and more patches were colonised than were abandoned. Patches that were continuously occupied (i.e. occupied during all sample periods) were typically of high quality (based on habitat and landscape characteristics), continuously unoccupied patches were usually of low quality, and intermediate quality patches were occupied intermittently. The variables explaining patch occupancy were similar each year, and a statistical model based on data from the year 2000 also predicted occupancy in 2004 with similar accuracy. However, data from a single survey were inadequate for identifying patches used intermittently by flying squirrels. Despite inconsistent occupancy, these patches may be important for the local persistence of flying squirrels. The dynamic occupancy pattern may thus affect estimates of suitable habitat area and identification of functional patch networks for landscape planning. These results emphasise the need for follow-up studies to better understand population patterns and processes in time.     

Colonization and extinction dynamics of an annual plant metapopulation in an urban environment

The metapopulation framework considers that the spatiotemporal distribution of organisms results from a balance between the colonization and extinction of populations in a suitable and discrete habitat network. Recent spatially realistic metapopulation models have allowed patch dynamics to be investigated in natural populations but such models have rarely been applied to plants. Using a simple urban fragmented population system in which favourable habitat can be easily mapped, we studied patch dynamics in the annual plant Crepis sancta (Asteraceae). Using stochastic patch occupancy models (SPOMs) and multi‐year occupancy data we dissected extinction and colonization patterns in our system. Overall, our data were consistent with two distinct metapopulation scenarios. A metapopulation (sensu stricto) dynamic in which colonization occurs over a short distance and extinction is lowered by nearby occupied patches (rescue effect) was found in a set of patches close to the city centre, while a propagule rain model in which colonization occurs from a large external population was most consistent with data from other networks. Overall, the study highlights the importance of external seed sources in urban patch dynamics. Our analysis emphasizes the fact that plant distributions are governed not only by habitat properties but also by the intrinsic properties of colonization and dispersal of species. The metapopulation approach provides a valuable tool for understanding how colonization and extinction shape occupancy patterns in highly fragmented plant populations. Finally, this study points to the potential utility of more complex plant metapopulation models than traditionally used for analysing ecological and evolutionary processes in natural metapopulations.     

Landscape characteristics associated with the occurrence of the flying squirrel Pteromys volans in old-growth forests of northern Finland

We investigated whether the occurrence of the flying squirrel, a declined virgin forest species, was accounted for by the structure of the landscape in northeastern Finland We sampled 20 forest areas (1–77 km² in size) in 1995 for the occurrence of the species using a sampling plot method Landscape structure around occupied and unoccupied areas were compared using seven different radii ranging from 100 m lo 10 km Classified satellite images were analyzed with Geographic Information System (GIS) Ten areas could be determined to be occupied by the flying squirrel Results showed that occupied areas did not deviate from unoccupied ones in total area In unoccupied areas there were more mature pine-spruce forests, and less matrix habitat at the home range scale (100 m and 200 m radii) than in occupied areas Further more, in unoccupied forest areas open habitats (clear-cuts, open fens) were more common than in occupied ones at distances between 200 m and 2 km around sampling plots These differences resulted from both larger mean patch size of and smaller mean nearest neighbor distance between open habitat patches m unoccupied than in occupied forest areas According to a logistic regression analysis, information on the amount of open habitat within 1 km radius alone correctly classified 75% of the areas into occupied and unoccupied ones The probability of the presence of flying squirrel decreases with the increasing amount of open habitat within 1 km The results suggest that increasing the amount of open habitats (e g clear-cutting) at the local scale has a negative impact on the flying squirrel It is obvious that in the present situation where the population has severely declined and its optimal habitat is still diminishing, every patch suitable for the species - whether occupied or not -may be important for the population persistence     

Conspecific Attraction is a Missing Component in Wildlife Habitat Modeling

Abstract Wildlife biologists use knowledge about wildlife-habitat relationships to create habitat models to predict species occurrence across a landscape. Researchers attribute limitations in predictive ability of a habitat model to data deficiencies, missing parameters, error introduced by specifications of the statistical model, and natural variation. Few wildlife biologists, however, have incorporated intra- and interspecific interactions (e.g., conspecific attraction, competition, predator-prey relationships) to increase predictive accuracy of habitat models. Based on our literature review and preliminary data analysis, conspecific attraction can be a primary factor influencing habitat selection in wildlife. Conspecific attraction can lead to clustered distributions of wildlife within available habitat, reducing the predictive ability of habitat models based on vegetative and geographic parameters alone. We suggest wildlife biologists consider incorporating a parameter in habitat models for the clustered distribution of individuals within available habitat and investigate the mechanisms leading to clustered distributions of species, especially conspecific attraction.     

Dispersal, distribution, patch network and metapopulation dynamics of the dingy skipper butterfly (Erynnis tages)

Two general approaches have usually been taken towards understanding the distributions and dynamics of localised species in heterogeneous landscapes, namely habitat characterisation and metapopulation dynamics. We show how habitat and metapopulation dynamics interact to generate a highly localised distribution of a butterfly, despite the extremely widespread nature of the butterfly’s host plant. Egg placement, macro-habitat requirements and dispersal were studied for the butterfly Erynnis tages, in North Wales, where it shows a restricted distribution relative to that of its host plant, Lotus corniculatus. Females laid eggs disproportionately on large plants growing in hollows, with intermediate cover of bare ground and high cover of L. corniculatus. Ideal macro-habitat, studied at 100-m grid resolution, consisted of areas with high host plant densities, sheltered from wind, with light or no grazing or cutting. These specialised conditions are represented as localised patches in the landscape, and define the potential habitat network, within which metapopulation dynamics take place. Although there was a moderate (22%) level of exchange of individual E. tages among local populations, the total number of potential colonists in the whole system was low because source population sizes were small (≤200 individuals at peak in any site in 1997 and 1998). Four unoccupied but apparently suitable 500-m grid squares were colonised between 1997 and 1998, and isolated habitat was less likely to be occupied. Overall, our study suggests that long-term regional persistence of E. tages is very likely to depend on metapopulation processes within the restricted patch network, rather than on the long-term survival of local populations. Received: 25 May 1999 / Accepted: 9 August 1999