Successful Conservation of a Threatened Maculinea Butterfly

See allHide authors and affiliations

Science  03 Jul 2009:
Vol. 325, Issue 5936, pp. 80-83
DOI: 10.1126/science.1175726


Globally threatened butterflies have prompted research-based approaches to insect conservation. Here, we describe the reversal of the decline of Maculinea arion (Large Blue), a charismatic specialist whose larvae parasitize Myrmica ant societies. M. arion larvae were more specialized than had previously been recognized, being adapted to a single host-ant species that inhabits a narrow niche in grassland. Inconspicuous changes in grazing and vegetation structure caused host ants to be replaced by similar but unsuitable congeners, explaining the extinction of European Maculinea populations. Once this problem was identified, UK ecosystems were perturbed appropriately, validating models predicting the recovery and subsequent dynamics of the butterfly and ants at 78 sites. The successful identification and reversal of the problem provides a paradigm for other insect conservation projects.

The conservation of insects poses formidable challenges (1). National extinction rates of temperate butterflies and other arthropods have recently exceeded those of terrestrial vertebrates and vascular plants (26), and population extinctions have frequently occurred on nature reserves where species’ resources remained abundant (68). Moreover, every early attempt to conserve a declining butterfly failed because of inadequate understanding of the causes of decline (7, 8). In 1974, the International Union for Conservation of Nature therefore selected three butterflies, including the ~six species of Maculinea (Large Blues), as global flagships for lepidopteran conservation (9), advocating research into their ecology and the maintenance of source habitats (10); the longest-running initiative involves Maculinea arion.

M. arion is an extreme specialist that switches from feeding on a plant to living as a social parasite inside Myrmica ant colonies during a 10-month larval instar and 3-week pupal period (Fig. 1). In the UK, M. arion’s population was estimated to include 91 colonies over the period of 1795 to the 1840s, declining to ~25 populations supporting tens of thousands of adults in 1950, and to 2 colonies of ~325 total individuals in 1972 before national extinction in 1979 (Fig. 2A) (11, 12). Nine sites were declared conservation areas from 1930 to 1969, which preserved M. arion’s Thymus- and Myrmica-rich grasslands but failed to slow extinctions (12).

Fig. 1

Life cycle of M. arion. Adult butterflies oviposit on Thymus species flowers from June through July (model parameter Vw). Larval instarsI-III feed on flowerheads for 3 weeks, with (including eggs) survival (SF) depending on parasites, predation, and cannibalism. The small final instarIV larva abandons Thymus and is adopted into the underground nest of the first Myrmica ant worker to encounter it, with survival (Pt) depending on its adoption into primary (M. sabuleti) or secondary (M. scabrinodis) host-species’ nests. M. arion larvae acquire ~98% of their final biomass eating ant brood and frequently experience density-dependent mortalities (1 – Dt) in nests that adopt more than 1 larva; mortalities in ant nests are amplified in drought years (QW). After 10 months, the larva pupates in the Myrmica nest, emerging as an adult 2 to 3 weeks later. [Illustrations by Richard Lewington]

Fig. 2

Populations of M. arion in the UK. (A) Number of colonies since 1950. Hatched bars indicate temporary colonies. Arrows: “a” represents extinction of native UK populations, “b” indicates re-introduction from Sweden to UK restoration sites, and “c” denotes introductions using UK populations. (B) The persistence of 39 UK populations known in 1929 to the extinction of the last in 1979, plotted against each population’s site area and the square root of distance from the nearest neighboring occupied habitat patch in 1929. No colonization was observed during this period. Red, Atlantic coast populations; blue, Cotswolds; black, Dartmoor. No relation was found between persistence time and either landscape parameter [areaallsites: R2 = 2%, P = 0.39; isolationallsites: R2 = 5%, P = 0.20; areawithinlandscape: R2 = 1 to 14%, P = 0.70 to 0.18; isolationwithinlandscape: R2 = 5 to 9%, P = 0.46 to 0.16].

To understand M. arion’s decline on superficially unchanged sites, annual variation in every factor causing mortality or reduced natality in the life cycle was identified and measured from 1972 to 1978 in its last UK population on Site X, Dartmoor (fig. S1) (13). We quantified 18 life-table parameters during 6 years of typical and extreme weather (table S1), including adult dispersal (14), oviposition choice and egg distribution (15), adult natality, egg and larval mortality on Thymus species, cannibalism on Thymus (16), adoption by Myrmica ant species (17), host specificity (18), queen effect on workers (19), and the carrying capacity of ant colonies (20). Mortality on Thymus (egg-larval instarIII) was low and had no influence on population dynamics (fig. S2), although the distribution of Thymus determined which Myrmica ant nests were accessible to larvae (17). Post-adoption larval instarIV mortality inside Myrmica nests was the key factor determining overall population changes (fig. S2), an analysis supported by phenomenological and spatial automata models (16, 21). We identified four causes of M. arion mortality at this stage: (i) availability of the preferred ant host species (18), (ii) presence/absence of queen ants (19), (iii) larval density within ant nests (20), and (iv) drought effects on host food availability (22). Factors (i) and (iii) were most variable: Five Myrmica species foraged beneath Thymus on Dartmoor and adopted instarIV larvae with probabilities proportional to worker abundance (17), but survival was 5.3 times higher in colonies of Myrmica sabuleti than with those of its congeners (18); population-scale survival with M. sabuleti decreased with larval density per nest (fig. S3).

Failure by female butterflies to lay their full potential of eggs amplified the population decline from 1974 to 1976 (fig. S2). We observed no emigration (14) or eggs on sites 260 to 400 m away from primary colonies. Predation by birds and invertebrates was high but stable, reducing the mean female longevity to 3.9 ± SE 0.6 (standard error of 0.6) days and natality to 51 ± 8.9 eggs, which is about 25% of the number laid by caged females. After single inactive days of wet weather, an average female laid 56 ± 14% more eggs on the next suitable day but failed to compensate for prolonged wet periods in 1974 (table S1). In contrast, a drought in 1975 reduced mean female life spans to 2.1 ± 0.3 days and average natality to 20 eggs. On the basis of historical records, there is only an 8% chance of two such extreme summers in a 7-year period, but both were included as model parameters due to the unpredictability of future weather.

We reduced the 18 life-table factors to seven influential parameters in a mechanistic model describing M. arion’s population dynamics on its last UK site in the period of 1972 to 1978at+1=atSF(Psab(t)+cPscab(t))DtQwVw(1)where at is the predicted egg numbers in year t; SF is the mean survival of eggs and early larval instars on Thymus [estimate = 0.37, see supporting online material (SOM)]; Psab(t) and Pscab(t) are the proportion of instarIV larvae adopted respectively by M. sabuleti or other secondary-host species in year t (larvae die if no Myrmica coincides with Thymus); c is the initial density-independent survival within secondary-host ant nests (0.19) (18); Qw is the reduction in larval survival in Myrmica nests in drought years (0.56) (table S1); Vw is the number of eggs per adult (table S1) in a typical year (25.5 eggs), a prolonged wet summer (11.5), or an extreme drought summer (8.0); and Dt is the density-dependent survival within Myrmica nests, whereDt=αβlog(Lt/Rt)(2)and Lt is the larval density per Myrmica nest in year t = (total larvae entering ant nests)/(number of available nests), and Rt is the relative worker size = (mean worker ant weight in year t)/(mean worker ant weight at Site X1972-77). The calibration data fit to Eq. 2 is shown in fig. S3.

We found that Eq. 1 fitted actual egg population numbers [R2 = 95% (where R is the correlation coefficient), P = 0.005] and annual changes (R2 = 76%, P < 0.05) on Site X from 1972 to 1978 and, thus, was used to estimate the amount of habitat manipulation needed to optimize conditions for host Myrmica species in all subsequent conservation projects. Varying the ratio of M. sabuleti [Psab(t)] to other Myrmica species [Pscab(t)] co-occurring with Thymus predicted a butterfly growth rate λ of greater than 1 if more than 68% of larvae are adopted by their primary host, assuming the low butterfly densities and conditions on Site X in the 1970s. In optimum habitat of 100% Thymus coexistence with host ants, no extreme weather, and the median ant densities and biomass subsequently recorded on conservation sites, we predicted λ = 5.49, stabilizing after 12 years under density-dependent parameter Dt at a capacity (K) of 910 adults per hectare.

Surveys and habitat analyses (13) of every former M. arion site in fig. S1 enabled us retrospectively to analyze the cause of its decline. Overexploitation (butterfly collecting) was ruled out as a factor (11, 12), and no relation was observed between M. arion population survival time and grassland patch size or isolation over the 50 years before national extinction (Fig. 2B). Instead, we found that the quality of grassland habitat, represented by the availability of M. sabuleti, was the driver to extinction. Despite abundant Thymus and other Myrmica species, only 0 to 38% of food plants coexisted with M. sabuleti on former sites (Fig. 3A), well below the minimum predicted for λ > 1. Only on Site X, where the butterfly persisted, did M. sabuleti densities exceed the 68% model threshold. M. sabuleti presence was closely correlated with vegetation structure; in particular, occurrence decreased with turf height (Fig. 3A and fig. S4B). The fitted relation suggested swards less than 1.4 cm tall were needed to provide the minimum threshold host-ant density for M. arion. Turf height, in turn, correlated with soil temperature (fig. S5) during key growth periods for M. sabuleti (23). When mean swards exceeded ~2 cm, the microclimate in their warmest daytime brood chambers near the soil surface fell more than 2° to 3°C relative to 1-cm tall turf, allowing less thermophilous congeners, especially M. scabrinodis, to outcompete M. sabuleti (fig. S4C) (23). As human mediated grazing declined, the hillsides to which M. sabuleti was confined were grazed primarily by rabbits. When the rabbit population declined because of myxomatosis, all but one (farmed) M. arion site became too overgrown for the primary host (Fig. 3A).

Fig. 3

Relation between the proportional occurrence (Psab) of ant M. sabuleti in Thymus grasslands and turf height (H). Horizontal dashed lines denote that minimum Psab (68%) was required for M. arion’s basic reproductive rate λ > 1. (A) Model calibration was calculated on the basis of Psab and H recorded from 1972 to 1978 surveys. Blue, occupied M. arion sites (dot at 0.42 denotes the site in year of extinction); red, known extinct sites; black, other sites. Fitted empirical logistic: loge[(Psab + 0.01)/(1 – Psab + 0.01)] = αH + βH (1/√H + 0.1), αH = –5.95 (SE = 0.40), βH = 8.12 (SE = 0.74), R2 = 56%, P < 0.001. (B) Annual relation after application of grazing regime on site Y from 1974 to 2008. Dots denote experimental grassland [colors as in (A)], and triangles represent unmanaged controls. (C) Model application to 77 other restoration sites sampled (N = 334) from 1978 to 2008. The dashed line indicates an empirical logistic fit to restored sites: αH = –6.63 (SE = 0.40), βH = 11.02 (SE = 0.64), R2 = 61%, P < 0.001.

Unaware of the inconspicuous habitat degradation, early conservationists took inappropriate measures on the basis of false assessments of M. arion’s decline. Thus, a fence erected in 1931 to deter butterfly collectors inadvertently excluded the herbivores that maintained its specialized habitat, resulting in rapid extinction of the population (12). By the time the cause of the decline was evident, the last UK population was extinct (Fig. 2A). Since then, 52 Thymus grasslands have been managed and restored, and at least 50 additional sites are at earlier stages of restoration, with the criteria defined by Eq. 1 (Fig. 3). Although isolation and fragmentation were not implicated in M. arion’s decline (Fig. 2B), where possible chains of sites were restored within the butterfly’s ~250-m dispersal range (14) in four landscapes (fig. S1). Meanwhile, a suitable donor race was identified from Öland, Sweden, with a phenology that synchronized with its resources under observed and predicted UK climates (fig. S6).

Myrmica populations responded rapidly to the changed vegetation structure. On Site Y, Dartmoor (fig. S1), M. sabuleti was not detected from 1973 to 1974, yet it coexisted with 84% of Thymus within a decade after scrub clearance and the reintroduction of seasonal grazing, while remaining absent from ungrazed controls (Fig. 3B; R2 = 81% for fit of niche-model). Similar patterns occurred on 77 other conservation grasslands (Fig. 3C), where host-ant co-occurrence with Thymus initially averaged 9% (45% maximum). On many managed sites, M. sabuleti now dominates 100% of Thymus-grassland, sometimes forming super colonies with 5 to 10 nest-centers per square meter and mean worker (hence brood) weights 13 to 82% higher than on Site X in the 1970s. Compared with early data, the fit with turf height was less precise, although recalibration showed the same strong pattern (Fig. 3C; R2 = 37 and 61%, respectively), with the minimum threshold host-ant density for M. arion now occurring in turf less than 2.1 cm tall. This tolerance of taller swards by M. sabuleti is largely explained by recent climate warming reducing the vegetation’s shading effect (24). Thymus populations were less dynamic. On average, the density of flowering plants increased by 265 ± 80% but dispersed to encompass previously inaccessible Myrmica colonies only at two sites.

Successful releases of M. arion (SOM) were made from Öland, Sweden, to three UK sites in 1983 to 1992 (a fourth to the Cotswolds failed, presumably due to an excessive climatic difference relative to Öland). Four later introductions (Fig. 2A) to distant sites or new regions were sourced from established populations in Somerset, UK. By 2008, the butterfly had naturally colonized 25 other conservation sites, albeit establishing peripheral colonies in seven cases (Fig. 2A). The largest populations contained 1000 to 5000 adult butterflies per hectare, an order of magnitude greater than any previous population of M. arion recorded worldwide.

Long-term predictions of butterfly population dynamics (Fig. 4A and figs. S7 and S8) were estimated from the initial number of M. arion (a0) introduced in year 0, and thereafter from annual variation in habitat and weather parameters Psab(t), Pscab(t), Dtw, and Vw, with no adjustment for observed butterfly numbers during its subsequent 16 to 21 generations. Observed M. arion populations closely followed model predictions (Fig. 4A and fig. S7), increasing in 5 to 10 generations toward their theoretical carrying capacities, where they stabilized or experienced temporary declines because of drought (e.g., 1996 to 1997) or suboptimal management (Site X, 2000 to 2004). In seven M. arion populations that have existed long enough for density-dependent mortalities to halt further growth, observed peak densities correlate closely with estimated site carrying capacities per square meter (Fig. 4B): The fact that site values vary by two orders of magnitude is due to differences in the ratio of M. sabuleti to other ants, Myrmica nest densities, and worker biomass. Because of the parameterization of stabilizing density-dependent larval mortalities within Myrmica nests (Eq. 2), model predictions of butterfly numbers 10 to 20 years ahead were as close to observed numbers as in the first decade (fig. S8). On more recently restored UK sites, the direction and size of annual M. arion population change was also close to model predictions (Fig. 4C).

Fig. 4

Prediction of the long-term population dynamics of M. arion in the UK. (A) Fit of Eq. 1 to recorded changes on Site X (blue lines), from which it was parameterized in the period of 1972 to 1978, and its ability to predict observed numbers after M. arion was re-introduced following conservation management to Site X (Embedded Image = 69%) (SOM) and Green Down (green lines, Embedded Image = 94%). Solid lines indicate observed numbers, dashed lines denote model predictions, and K indicates the site carrying capacity (equilibrium level) for M. arion estimated from Eq. 1, where density effects (Dt) negate further growth. (B) Estimated and observed values of carrying capacity per square meter. Dotted line, 1:1 ratio; solid line, fitted regression; R2 = 98.6%; P < 0.01 (before conservation management, K = 0 except on Site X). (C) Prediction of annual changes (N = 99) in M. arion egg numbers on 19 conservation sites from 1983 to 2008 (red, Site X). The plot shows the log of observed change [log(Et+1/Et)] versus the log of predicted change [Embedded Image], where Et is the observed egg number in year t, and Embedded Image is the predicted egg number in year t + 1 from Eq. 1 with at replaced by Et. Dotted line, 1:1 ratio; solid line, fitted regression; R2 = 71%; P < 0.001.

M. arion’s UK re-establishment is too recent for settled landscape-scale dynamics to emerge. During its natural colonization of 23 Polden Hills’ sites, M. arion spread by stepping-stone occupation of neighboring habitat patches, taking 14 years to reach the furthest one, 4.4 km from a source introduction. The rate of increase of colonization distance from original sources was 1.9 times greater during years 8 to 14 than in years 0 to 4 (P < 0.02), possibly reflecting selection for dispersive adults consistent with shifts in thorax widths measured in UK M. arion populations during previous range changes (25). Seven sites experienced temporary occupancy in 2003 to 2008 (Fig. 2A): All were smaller than other conservation areas (average of 0.11 and 1.55 ha, respectively, and Mann-Whitney P = 0.008) and are regarded as ephemeral satellites of neighboring populations.

Surveys of M. arion across Europe have revealed similar declines for similar reasons (SOM). Today, successful but smaller-scale management is being applied in other nations on the basis of UK results, and M. arion’s global listing has changed from “vulnerable” to “near-threatened” (26). It will shortly downgrade from “endangered” to “vulnerable” in Europe and is one of just three UK butterflies due to meet the Convention of Biological Diversity’s target to reverse species’ declines by 2010 (27). The others inhabit similar-structured grassland: Their successful conservation derived directly from extensions of this research (7).

The Maculinea project tackled problems typical of many temperate butterflies that were disappearing from apparently suitable sites (7, 8) and provided insights for quicker, cheaper approaches. Having recognized that immature stages (and ant nest sites) typically exploit a narrow subset of their named resource and that the availability of optimum larval habitat alongside adult metapopulation constraints largely determines population sizes and persistence (7, 28), successful conservation was achieved across the genus Maculinea. Similarly, the declines of several phytophagous Palaearctic butterflies were reversed by identifying the subsets of food plants preferred by larvae, then perturbing ecosystems to generate them (7). Despite these short-cuts, species-conservation is impractical for the vast majority of insects, for which the preservation of primary ecosystems or a community approach are appropriate (1). However, because many other threatened species increased on UK sites following targeted management for M. arion (29), we consider that successful species-based and community conservation for insects represent different routes to the same end.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S8

Table S1

References and Notes

References and Notes

  1. Materials and methods, as well as additional information, are available as supporting material on Science Online.
  2. We thank R. M. May for comments, R. Lewington for Fig. 1 illustrations, and 140 others for contributions and assistance (see reference S16). Funding, experimental sites and other resources were generously provided by the European Union Macman and Biodiversa (CLIMIT) programs, Natural England, Centre for Ecology and Hydrology, the National Trust, the Somerset Wildlife Trust, Network Rail, J and F Clark Trust, Butterfly Conservation, Gloucester Wildlife Trust, Millfield School, Defra, the World Wildlife Fund, Sir Terence Conran, Holland and Barrett, Hydrex, ICI, and R. Mattoni.
View Abstract

Navigate This Article