Tomas Roslin, Metapopulation Research Group, Department of Biological and Environmental Sciences, University of Helsinki, PO Box 65 (Viikinkaari 1), FI-00014 Helsinki, Finland. E-mail: email@example.com
Abstract. 1. Habitat loss and fragmentation is a leading cause of species extinction. This not only concerns loss of major habitats, but also loss of microclimatic heterogeneity within such habitats.
2. In this study, we examine the effects of microclimate on the abundance and movements of Onthophagus gibbulus, a dung beetle associated with pastoral habitats. While formerly widespread in Southern Finland, the species is now critically endangered at a national level, persisting within an area of 6 km2.
3. We divided the Finnish distribution of O. gibbulus into 50 × 50-m grid cells, characterised local microclimate by incident solar radiation, and surveyed the distribution of O. gibbulus within a subset of cells. To investigate the impact of microclimatic conditions on dung beetle movements, we conducted a mark–release–recapture study.
4. Our approach allowed us to estimate the national population size of O. gibbulus– a figure rarely available for endangered insect species. The Finnish population of O. gibbulus comprises between 2000 and 6000 individuals.
5. The abundance of beetles per dung pat increases with incident solar radiation, and beetles are more likely to move towards warmer than colder spots in the landscape.
6. Overall, our study depicts O. gibbulus as a thermophile confined to the warmest part of the landscape, and offers loss of microclimatic variation as the cause for its large-scale decline. To conserve O. gibbulus and similar species, we need to consider not only the amount of macrohabitats, but also the amount and distribution of microclimatic variation within them. Future predictions are complicated by ongoing climate change.
Throughout the world, habitat loss and fragmentation is a leading cause of species loss. This pattern not only relates to the disappearance and fragmentation of major habitats, but also to an even quicker loss of microhabitats within them (Hanski, 2005). As the disappearance of microhabitats is more difficult to quantify than the loss of major habitats, it has received less emphasis in global conservation policies. Nonetheless, a general decrease in environmental heterogeneity may lead to reduced buffering against environmental extremes, thereby increasing the risk of extinction even for generalist species (Kindvall, 1996; Ehrlich and Hanski, 2004; Hanski, 2005).
For specialist species, the loss of particular microhabitats may result in more specific effects. Such species will experience a patchy landscape, with patches defined by combinations of specific environmental conditions (MacArthur and Levins, 1964; van Nouhuys, 2005). Here the spatial configuration of environmental variation may have a prime effect on population patterns and processes (Hanski, 1999a; Heikkinen et al., 2005).
Insect species specialised on agricultural habitats provide interesting study systems for exploring how habitat loss and modification affect animal movement and population dynamics. Insects are often specialised on particular microhabitats, especially when close to their range margin, posing a particular challenge to insect conservation (e.g. Thomas, 1991; Thomas, 1995; Bourn and Thomas, 2002; Samways, 2005; Stewart et al., 2007). Early successional habitats generated by agricultural practicies offer some particularly warm microclimates, sustaining a substantial fraction of e.g. British butterflies (Thomas, 1993; Thomas et al., 1999). In much of Northern Europe, macrohabitats associated with traditional agriculture have declined by more than 90% during the 20th century (Pykälä and Alanen, 2004). For microhabitats, we can only guess at current loss rates – but they are bound to be higher (cf. Cowley et al., 1999; Tiainen et al., 2004).
In this context, dung beetles in the genus Onthophagus are of particular concern. Of the three Onthophagus species formerly encountered in Finland, two are currently nationally extinct (Biström et al., 1991; Rassi et al., 2001; Roslin and Heliövaara, 2007). Yet, one species –Onthophagus gibbulus– still persists. While formerly widespread in Southern Finland, the current distribution of the species is restricted to a single area of ∼6 km2. Just why the population has survived here (and why it has disappeared elsewhere) remains unknown. As the current distribution of the species encompasses an area of exceptional variation in microhabitats, we hypothesised that the species has retracted to the only parts of the landscape still offering suitable microclimatic conditions. If so, we would expect the local abundances of O. gibbulus to covary with local environmental conditions, and the movements of individuals to follow microclimatic gradients in the landscape.
In this paper, we use the remnant population of O. gibbulus as a model system to examine how microclimatic variation may affect animal abundance and movement in disappearing habitats. Operationally, we define microclimatic variation as the amount of solar radiation intercepted by the ground, and test the explicit hypotheses that the current distribution of O. gibbulus is confined to the warmest or coldest parts of the landscape, and that movements of individuals are canalised by local microclimatic gradients. Finally, we derive an estimate of the current population size of the species, and speculate about how this and similar species might best be managed in our ever-changing rural landscapes.
Material and methods
To examine the effects of microclimate on the movements and abundance of O. gibbulus, we combine two approaches: an observational survey of the local abundance of the species in natural dung pats, and an intensive mark–release–recapture study within the area sustaining the stronghold of the population.
Onthophagus gibbulus is a brownish beetle ranging in size from 0.8 to 1.5 cm (Ljungberg, 2002; Fig. 1). Typical individuals are easy to sex, with the males characterised by a distinct horn in their neck, and the females by a prominent protrusion at the anterior end of their pronotum (Roslin and Heliövaara, 2007). The species is diurnal, and lays its eggs in tunnels dug under the dung pat. Tunnels typically include several brood chambers provisioned with dung reserves (Burmeister, 1930, 1936). Part of the population seems to hibernate as adults, since fullgrown beetles can be encountered throughout the snow-free part of the year. Nevertheless, population densities reach a marked peak in August and September (Roslin and Heliövaara, 2007; Avomaa and Kataja unpublished), which is when this study was conducted.
Onthophagus gibbulus is currently classified as critically endangered in Finland (Rassi et al., 2001). Between 1961 and 1995, no records were made of the species, and it was already considered regionally extinct (Biström et al., 1991). In 1995, a remnant population was found in the south-western part of Finland, in the Häntälä region near Somero (60°34′N, 23°22′E). While the species has later been actively searched for, no new occurrences have been found in any other parts of the country (Roslin and Heliövaara, 2007).
The current distribution of O. gibbulus is confined to the upper course of the Rekijoki River. The region is characterised by broad and flat expanses of intensively cultivated clay soils. A detailed description of the area is given by Luoto (2000) and by Kontula et al. (2000). In brief, the river valley is marked by steep meadow slopes exposed to different directions. Many of these slopes are grazed, and overall, the region sustains an unusual range of habitats typical of traditional agriculture. At present, Rekijoki is the largest remaining area of mesic grasslands in Finland (Kontula et al., 2000).
Outside of Finland, the closest populations of O. gibbulus occur in the Baltic countries and in the Moscow region. The general range of the species is continental, with its centre of gravity in Eastern Europe and Siberia (Horion, 1958; Bunalski, 1999; Roslin and Heliövaara, 2007). The westernmost populations occur in Poland and Italy, whereas in the East, the range extends to Asia Minor, Mongolia, Eastern Siberia and North Korea. As for most endangered arthropods, information on population trends from other parts of the range is not readily accessible.
To examine microclimatic factors affecting the abundance and distribution of O. gibbulus, we delineated its distributional area as a continuous 6.02 km2 block (Fig. 1). To describe microclimatic variation within the region, we split the larger block into 2 408 50 × 50 m grid cells, and used variation in thermal conditions resulting from differences in incident solar radiation on slopes of different exposure to characterise microclimatic variation. For this purpose, we adopted the computer model of Luoto et al. (2001; original method from Griffiths, 1985) to calculate maximum theoretical solar radiation for each grid cell as a ratio between 0 and 100:
where altitude is the angle of sun above horizon; sun direction is the direction of the sun with respect to the South (increasing towards the West: South = 0, East = −90, West = +90); ground direction refers to the direction of the slope of the ground (South = 0, East = −90, West = +90); and slope angle refers to the steepness of the slope of the ground (in degrees: a flat area = 0, a vertical slope = 90). Solar radiation was calculated as the average of radiation during three moments during the diurnal activity period of O. gibbulus: 09.00, 12.00 and 15.00 hours.
The mark–release–recapture work was focused on the central part of the study area (Fig. 1). Here we selected 32 grid cells within a continuous pasture of ∼15 ha. The cells were chosen so that they represented the overall distribution of solar radiation in the region (mean 71.5, SD 3.7, range = 64–79, n = 32 for the selected cells vs. mean 70.0, SD 3.2, range = 55–79, n = 2408 for the full study area), while still forming a continuous field to maximise the detection of marked individuals. All cells were selected on the northern shore of the river, since the water largely confined grazing cattle to this part of the pasture. During the study period, the trapping area was grazed by a heard of 20–30 heifers.
Following Roslin (2000), each trap consisted of a tinplate funnel with an upper diameter of 35 cm, a lower diameter of 4 cm, and a depth of 25 cm. The funnel was buried in the ground to its upper brim. A cow pat of standard size (∼2 kg fresh weight) was placed on a metal grid (mesh size 2.5 cm) on top of the funnel. The bait pats were made of thoroughly mixed fresh cow dung collected in a cow barn during the same morning. To keep the baits attractive, we used a sharp knife to peel the hard surface off the pat 2 days after exposure.
Beetles attracted by the bait fell into a 1-L plastic bottle. This bottle was filled with egg cartons, to offer the beetles a place to hide and thereby reduce the probability of escape. Each trap was checked twice per day, and any Onthopgahus encountered was marked by a code consisting of three small holes in different positions on the elytra (cf. Unruh and Chauvin, 1993; Nilsson 1997a, b; Roslin 2000). The holes were made with the tip of the thinnest available insect pin, and could only be seen under a microscope with a magnification of 6–12×. Each beetle received an individual code.
Immediately after marking, the beetles were transferred to a small piece of dung placed outside of the trap funnel. To make the beetles resettle in the dung, beetles to be released were put in small plastic tubes, which were then inserted in the dung pat with their open end down. Hence, the beetles could only exit the tube by digging into the dung pat. No beetles were seen to fly off immediately after the release.
The mark–release–recapture study was scheduled according to the hierarchical method of Pollock (1982). Here, the rationale is to split the sampling effort so that estimates of population size can be derived from several complementary types of information: the relation between numbers of marked and unmarked individuals (Schnabel, 1938), and information on the time when a marked individual was last seen (Jolly, 1965). We conducted six sampling periods of 4 days each, as separated by breaks of at least 4 days (3–6.VIII; 11–14.VIII; 19–22.VIII; 29.VIII–3.IX; 11–15.IX and 26–29.IX, 2004). The breaks reduced the total time spent by beetles in traps and ensured each individual a chance to reproduce in natural dung pats, thereby lessening the potential impact of the experiment on the endangered target species.
Between August 10 and September 26, 2004, 55 natural cattle dung pats were sampled within a subset of grid cells (Fig. 1). The choice of cells was determined by practical considerations, since pastures in this region are largely confined to the river valleys. Beetles were extracted by flotation of the pats in buckets of water. Koskela and Hanski (1977) have found this method to retrieve more than 95% of the beetles present in a dung pat. As cattle droppings attract maximum numbers of dung beetles a few days after deposition, we carefully selected pats with the physical appearance typical of that stage.
We note that compared to the baited traps used in the mark–release–recapture study, the sampling of natural dung pats will typically yield lower numbers of dung beetles. Among multiple reasons, two stand out: First, the number of beetles present in a dung pat will represent a balance between continuous immigration and emigration of beetles, whereas traps will exclude emigration and retain all immigrants (e.g. Finn and Giller 2000). Second, the baits used in our mark–release–recapture study were kept in a continuously fresh state by regular change and peeling (cf. above), whereas natural dung pats will only attract beetles for a relatively short period of time (Hanski, 1980).
Effect of microclimate on Onthophagus abundance. Drawing on the material from 55 natural dung pats, we modelled the pat-specific count of Onthophagus individuals as a function of solar radiation in the surrounding 50 × 50-m grid cell, assuming a log link and Poisson distributed errors. Since observational data of this type will typically include variation beyond that captured by the model (such as uncontrollable variation in the age and physical characteristics of dung pats), we assumed the data to be overdispersed, and adjusted both standard errors and likelihood ratio statistics by the residual deviance as divided by the degrees of freedom. The model was fitted by sas for Windows version 8.02 (Proc Genmod; SAS Institute Inc., Cary, NC, USA).
Effect of microclimate on Onthophagus movement. The mark–release–recapture material was used to analyse how the characteristics of the individual (its gender), the distances between traps, and microclimatic gradients (defined as the difference in solar radiation between two traps) affected the observed trajectories of marked individuals. However, the movement data are also affected by the study design, which needs to be accounted for to avoid two potential biases. First, the closer to the end of the experiment an individual is marked, the shorter is the time available for recapture. Second, due to occasional rains and the risk of traps being flooded, some traps in our mark–release–recapture experiment were exempted from operation for given periods. Such spatio-temporal variation in trapping intensity will clearly affect the probability of detecting a given individual moving between given traps. Therefore, we first calculated the probability bjt with which an individual released at time t would be recaptured in a given trap j at any time after the release, assuming that neither gender, distance or solar radiation would have any effect on its movement patterns (but incorporating sex-linked differences in temporal recapture rates; cf. below). This probability can be calculated as
where cjt equals 1 for traps operational at time t and zero for all other traps, and P(t′–t) refers to the probability of recapturing an individual t′–t time steps after release.
We assume that P(t′–t) depends only on the length of the time lag t′–t (henceforth Δt), not on absolute time t. We let y(Δt) be the total number of individuals observed after time Δt following release, and z(Δt) the total number of traps in operation after time Δt (both summed over all release events). The overall probability of recapturing an individual after Δt is hence given by
To smoothen the raw data, we defined P(1) as p(1), and then fitted a power function p(Δt) = u(Δt − 1)v to the data for all Δt ≥ 2. Here, u refers to the probability of recapturing an individual after time lag Δt = 2, whereas ν is a scale parameter determining the rate of decline of p with Δt. The function p(Δt) was fitted to the data by maximum likelihood. To allow for sex-linked differences in temporal recapture rates, we fitted a separate curve to data on males and females. Eight individuals of uncertain gender were omitted from both this and later analyses.
Having derived the null expectation for movement patterns, we next tested whether realised patterns were modified by gender, distance and/or microlimatic gradients. To this end, we used variable bijt as our as response, scoring a value of 1 if an individual that was marked and released at trap i at time t was observed to move to a target trap j, and a value 0 if this was not the case. We modelled bijt with logistic regression as
Here, SameTrap refers to whether the movement occurred between the small ‘release pat’ next to the very same trap or between two distinct traps. Distance is the Euclidean distance between two traps (in m). Sex identifies the gender of the individual. Gradient is defined as the difference in solar radiation between traps i and j. Finally, the term logit(bjt) refers to the logit-transformed null expectation for observing a movement to any trap j at time t, as derived above. This term was entered as an offset variable, i.e. a term with a β-coefficient of 1. If none of the terms had any effect on realised movement patterns, all coefficients β1...9 would attain the value 0, and the model reduce to P(bijt = 1) = bjt.
The model was fitted with sas for Windows version 8.02 (Proc Genmod; SAS Institute Inc., Cary, NC, USA). Deviations from H0: β1...9 = 0 were examined by type 3 analysis of likelihood ratios. To derive quantitative estimates of movement rates, we used a model with non-significant factors removed.
During our examination of 55 natural dung pats, we found a total of 51 Onthophagus individuals in seven different pats. Despite sustantial variation in the quality of cow pats sampled in the field, solar radiation had a clear and significant imprint on the local abundance O. gibbulus: the higher the incident radiation in the surrounding 50 × 50-m grid cell, the higher the abundance of Onthophagus in a pat (χ21 = 20.54, P < 0.0001; Figs 1 and 2).
We caught and marked a total of 1339 Onthophagus individuals during the mark–release–recapture study (Fig. 3a), 140 of which were later recaught once, 21 individuals twice and two individuals thrice. The average distance moved between two traps was 140 m (SD 99 m).
Males and females differed in the behaviour after marking and release. In both sexes, the recapture probability was very low 1 day after the release, whereafter there was a distinct peak in male recaptures. In females, the corresponding peak was lower and less marked (Fig. 4a). Of all 188 recaptures, 51% related to movements between distinct traps, whereas 49% occurred between the small ‘release pat’ and the trap next to it (a pattern captured by factor SameTrap; Table 1). The gender of an individual did not affect the probability with which it moved between two separate traps (Table 1a: interaction SameTrap × Sex).
Table 1. Generalised linear model of factors affecting the movement of marked individuals of Onthophagus gibbulus. Table1A provides χ2 statistics for a type 3 analysis of likelihood ratios. Table 1B show regression coefficients measuring the effect sizes (with associated standard errors; SE) as derived from a model with non-significant factors removed. A)
Same Trap × Sex
Distance × Sex
Distance × Gradient
Gradient × Sex
Distance × Gradient × Sex
Distance × Gradient
Where the beetles moved to was affected by microclimatic variation. The probability of movement decreased with increasing distance between the two traps (Table 1: main effect of Distance), and the beetles tended to move towards warmer parts of the landscape (Table 1: main effect of Gradient). The effect of the gradient in solar radiation between two traps was stronger for movements over short than long distances (Table 1: interaction Distance × Gradient). Below a distance of 100 m, the probability of moving to a site with 10 U higher solar radiation was at least twice as high as moving to a site with the same amount of solar radiation as the release site (Fig. 4b). This dependence of movements on distance and solar radiation did not differ detectably among sexes (cf. Table 1, interactions Sex × Gradient, Sex × Distance and Sex × Distance × Gradient).
Estimates of population size obtained by the Jolly–Seber method and by Schnabel's method both showed a phenological pattern in the abundance of adult O. gibbulus (Fig. 3b). In late August, new individuals joined the population at a high rate (Fig. 3b, dashed grey line), resulting in a phenological peak in population size by the turn of August–September (Fig. 3b, black lines). During this maximal phase, the population size was estimated at 3350 to 6080 individuals with Schnabel's and the Jolly–Seber method, respectively (Fig. 3b). In early September, the recruitment rate already declined (Fig. 3b, dashed grey line). Among trapping periods, survival stayed high throughout the mark–release–recapture study (average 74%, range 51–100% between periods).
Our observations depict O. gibbulus as a thermophile confined to the warmest parts of the landscape. Where it occurs, its movements are canalised by fine-scale microclimatic variation. As the only remaining population of this species in Finland occurs in an area offering some exceptionally warm microclimates, our results suggest that the loss of microclimatic variation may be the cause behind the large-scale decline of this species in Finland.
Microclimate as a determinant of the abundance and distribution of Onthophagus gibbulus
With the intensification of European agriculture, the rural landscapes of Finland have fared no better than similar habitats elsewhere. Overall, the amount of Finnish meadows and pastures has declined from 62% of all agricultural areas to less than 1% today (Vainio et al., 2001). At a species level, the loss of agricultural habitats is directly reflected in Red Data Books: a rough third of the nation's endagerered species are currently associated with habitats emanating from traditional agriculture (Rassi et al., 2001).
In this context, O. gibbulus seems to offer a direct illustration of how the escalating loss of microclimatic variation within disappearing macrohabitats may affect the abundance and distribution of a species. From a scattered incidence across Southern Finland, the distribution of O. gibbulus has now retracted to a small area of a few kilometre squares only. Our results show that within this area, the abundance and movements of the species depend on microclimatic conditions: local abundances are highest in the warmest parts of the landscape, and movements of individuals occur along microclimatic gradients. Since the species is lacking from several pastures within its dispersal range (compare Figs 4b and 1), the restricted distribution of the species cannot be explained by the distribution of its macrohabitat, or of basic food resources (cattle dung). This reinforces the general notion that for a specialist species, the amount and distribution of macrohabitats may offer a poor description of the landscape as actually perceived by the species (MacArthur and Levins, 1964; van Nouhuys, 2005), and that of the landscapes of today, many insect species will occupy only the tiniest fraction (Cowley et al., 1999).
Onthophagus movement in a heterogeneous landscape
The results from our mark–release–recapture study suggest that the direction of dung beetle movements is affected by microclimatic conditions. At a qualitative level, a beetle will typically move from a colder part of a landscape towards a warmer part, but rarely the opposite way. As illustrated by Fig. 4b, the effect of the gradient is much more marked for short-distance movements than for long-distance movements. To put the observed effect size in perspective, we note that a difference of 10 U in solar radiation (assumed in Fig. 4b) amounts to less than half (42%) of the variation in solar radiation values observed across the landscape (range 55–79; n = 2 408 grid squares). Moreover, microclimatic variation occurs at a relatively fine scale, and a scale much smaller than the distances typically moved by dung beetles: within our study area, adjacent grid squares will frequently differ drastically in radiation values (Fig. 1). Hence, we expect that patterns of variation present in the landscape will have a major impact on realised patterns of dung beetle movements.
The model behind Fig. 4b is not based on mechanistic assumptions on movement behaviour, but is simply a statistical description of the movement data. As it does not separate the effects of the species movement behaviour from the design of the mark–recapture study, the quantitative results obtained here are specific to the the spatio-temporal intensity of trapping used in this study and should not be interpreted e.g. as the intrinsic dispersal kernel of the species (cf. Ovaskainen et al., 2008a). Nevertheless, the qualitative effects uncovered in O. gibbulus come with broader implications for studies of animal movement. One approach to modelling insect movements (as also applied to dung beetle movements; Arellano et al., 2008) is to divide the landscape into a number of habitat types, and to assume that the parameters of movement differ among these types (Ovaskainen et al., 2008b). While such models show predictive power at the landscape scale (Ovaskainen et al., 2008a), our results suggest that for some species, the effect of microclimatic variation may be so important that a model based on major habitat types alone provides an oversimplification. Another approach to modelling animal movement in heterogeneous space is to assume that the individuals track gradients in habitat suitability, prey densitiy, or other such variables (Grünbaum, 1998; Okubo and Levin, 2001). In the present case, the movements of the dung beetles clearly depended both on the distribution of the major habitats (pastures grazed by cows), and on microclimatic variables (temperature gradients) within the habitat. Hence, to build a model that would adequately capture the movement behaviour of O. gibbulus at the landscape scale, a combination of the two modelling approaches would be needed.
Managing the Onthophagus population
Our mark–release–recapture study focused on a subset of the whole population. Nevertheless, two considerations suggest that our estimates of population size represent the main part of the whole population: First, our survey of natural dung pats throughout the region identified the site used for mark–recapture as the population stronghold (Fig. 1). Second, estimated survival rates were high, averaging 74% between trapping periods. Since these estimates actually combine the probability of staying alive and of staying within the study area, we may infer that the population within this stronghold is relatively closed, and that high rates of recruitment evident in Fig. 3 largely reflect individuals hatching from local pupae. Even though specific estimates of peak population size differ by a factor of nearly two among the two methods applied, and while the associated confidence limits were relatively wide, the order of magnitude seems firmly established in the thousands (not in the hundreds or in the tens of thousands; cf. Fig. 3b).
A population size in the thousands will clearly exceed the most simplistic ‘rules of thumb’ for an imminent risk of extinction due to demographic or genetic stochasticity (e.g. Boyce, 1992). Yet, the Finnish population of O. gibbulus has rightfully been classified as critically endagered. It is restricted to a small area, and there is hence a high risk that a single perturbation might affect the full population. One such perturbation would be a break in habitat continuity (cf. Hanski, 1999b; Siitonen and Saaristo, 2000; Fritz et al., 2008). The pasture sustaining the main part of the population has been continuously grazed for hundreds of years (Kontula et al., 2000) – should grazing be discontinued for a single year, this might cause the population to crash.
From a management perspective, dung beetles like O. gibbulus offer interesting challenges: they are dependent on a resource (cattle dung) and habitat (warm, grazed slopes) both related to human activities. While national legislation will often prevent the active destruction of natural habitats inhabited by endangered species, forcing land owners to maintain vital resources for O. gibulus will seem like a complex task. Yet, we stress that the predicament of O. gibbulus is shared by a large number of taxa (Thomas, 1993), posing a challenge which we can hardly neglect. In this context, our results can be used to generate recommendations for the management of O. gibbulus, with ramifications for multiple taxa in changing agricultural habitats. First, since our results show O. gibbulus to be a specialist species confined to specific habitats within a small area, ensuring habitat continuity is a key priority. In this context, we recommend positive economic incentives to encourage local land owners to keep doing what they have done for centuries. Second, given the importance of microclimate revealed by our study, our model of solar radiation might be used as a tool in habitat restoration. By identifying those hot spots in the landscape which will offer the warmest conditions after cutting and grazing, we may maximise the reward of such conservation measures. Finally, our model of solar radiation could be used to scan other regions within Southern Finland for conditions suitable for O. gibbulus. Given dispersal limitation, such sites might offer promising sites for population re-introductions. In this context, we should end with a caveat: clearly, the microclimates of today will not be the microclimates of tomorrow. With globally rising temperatures, there is increasing evidence that several insect species are currently broadening their use of habitat types near their northern distribution limits (Thomas et al., 2001; Roy and Thomas, 2003; Davies et al., 2006). While at their range margins, these species have formerly been compensating for a cool macroclimate by selecting exceptionally warm microhabitats, closer to the centre of their range they have occupied a broader range of microhabitats (Thomas, 1993, Thomas et al., 1999). For such species, climate change might now increase the breadth of habitat use towards high latitudes, hence increasing habitat availability and population sizes, and decreasing dispersal distances between patches of suitable habitat (Thomas et al., 1999; Thomas et al., 2001, Davies et al., 2006). In addition to summer-time effects, warmer winter temperatures may facilitate survival in a wider range of microhabitats (Virtanen et al., 1998; Crozier, 2004). Thus, from the perspective of a thermophile like O. gibbulus, a wider part of the landscape may soon become available, in which case the remnant population of today would become an important source for future expansion (Thomas, 1993). Long-range dispersal from populations beyond Finnish borders will seem both slower and less likely. This underscores the need for conserving the current population of O. gibbulus in Finland – and those of similar species.
This study would not have been possible without the support of local landowners and cattle farmers of the Häntälä area, in particular the Oksanen family. Henna Seppälä and her family offered accommodation and invaluable local knowhow. Funding from the Finnish Ministry of Environment and the Academy of Finland (grant numbers 111704, 124242 and 213457 to TR and OO), as well as grants from Societas Entomologica Fennica and Societas Biologica Fennica Vanamo (to T.A. and M.K.) are gratefully acknowledged. Kari Heliövaara kindly supplied the species artwork for Fig. 1.