Journal of Insect Conservation

, Volume 15, Issue 1, pp 241–258

Evidence based conservation of butterflies

Authors

    • Department of ZoologyUniversity of Oxford
    • CEH Wallingford
  • D. J. Simcox
    • CEH Wallingford
  • T. Hovestadt
    • Department of ZoologyUniversity of Oxford
    • Muséum National d’Histoire Naturelle
Original Paper

DOI: 10.1007/s10841-010-9341-z

Cite this article as:
Thomas, J.A., Simcox, D.J. & Hovestadt, T. J Insect Conserv (2011) 15: 241. doi:10.1007/s10841-010-9341-z

Abstract

Few results of research aimed at solving questions arising from butterfly conservation are rigorously tested by manipulating populations and habitats in the field. Some factors common to successful conservation projects are analysed. In most non-migratory species, population density may vary by up to two orders of magnitude between sites or over time, and is primarily determined by the extent to which a subset of each species’ foodplant (or ant host) exists in the optimum growth-form or micro-habitat preferred by its larvae. Successful conservation projects have identified the optimum subset of each species’ larval resource before managing sites to increase its representation. In contrast, short-term fluctuations around a site’s carrying capacity or equilibrium level are mainly attributable to variation in weather, and are generally two orders of magnitude smaller than that attributable to larval habitat quality. There is little evidence that changing the abundance of adult resources, apart from shelter, influences population size or trends. The main constraint of the adult stage is the inability of many species to track the generation of new habitat patches that arise across modern landscapes. Within-patch larval habitat quality is again critical at the meta-population scale, explaining slightly more examples of patch occupancy than site isolation. This is because the higher density populations supported by optimum habitat are less likely to go extinct, and immigrants to new high-quality patches have a higher probability of founding new populations. Such patches may also generate up to a hundred times more emigrants per hectare than low-quality source patches.

Keywords

Habitat qualityMetapopulationButterfly population dynamicsLarval nicheDispersal

Introduction

The first scientific efforts to assess the changing status of butterflies, and to attempt to conserve species through managing habitats in accordance with research results describing their ecology and behaviour, began in the UK in the late 1960s (Pyle 1976). Since then, there has been an extraordinary expansion of monitoring and conservation-orientated research across Europe, North America, Japan and elsewhere.

From the outset, scientists have been more competent at describing the changes that were occurring in species’ distributions and abundance (e.g. in UK: Heath et al. 1984; Pollard and Yates 1993; Asher et al. 2001; Thomas 2005; Fox et al. 2006) than in understanding the mechanisms driving those changes; and despite some notable exceptions, the track record of reversing trends in declining species through targeted conservation is disappointing (see Thomas 1984a, 1991, 1995; Warren 1992; Pullin 1995; New et al. 1995; New 1997; van Swaay and Warren 1999; Kűhn et al. 2005; Settele et al. 2005, 2009; Thomas et al. 2009; Dennis 2010). Yet there is some cause for optimism. For example, by the late 1980 s the populations of six declining species had largely or entirely disappeared from UK nature reserves, several of which had been established mainly to conserve them (Thomas 1991): today four of these species (Maculinea arion, Polyommatus bellargus, Hesperia comma, Melitaea athalia) have mostly or wholly returned (or been re-introduced) following enlightened conservation management (Table 1), and with the exception of the guild of butterflies that inhabits early seral stages on woodland floors, several other rare specialists have increased or maintained status disproportionately well in conservation areas and, increasingly, on land managed primarily for forestry or under agri-environmental schemes (e.g. Brereton et al. 2008; Thomas and Lewington 2010).
Table 1

Recent status of populations of rare butterfly species on UK nature reserves and conservation areas compared with the period 1960–1989 (from Thomas 1991; Asher et al. 2001; Fox et al. 2006; Thomas and Lewington 2010)

Species

% Loss 1960–1989

2000–2009 Status

Lycaena dispar

100

Still extinct

Melitaea cinxia

0

100% present

Maculinea arion

100

25% return + >30 new

Polyommatus bellargus

27

100% return + many new

Hesperia comma

24

100% return + many new

Thymelicus action

0

10–15% range and population decline

Satyrium pruni

0

100% present + several new

Carterocephalus palaemon

100 (England only)

Still extinct

Melitaea athalia

100

100% return + net increases

The approach taken by conservation scientists has shifted over 40 years. At the outset, intensive ecological studies were made of 10 European species, involving the compilation of sufficiently detailed life-tables—and the identification and quantification of every factor causing mortality or reduced natality—for a key factor analysis to be made (see Thomas 1984a, 1991; Thomas et al. 1998a for case histories and further references). Although, to our knowledge, these studies generated the only datasets from which mechanistic, as opposed to correlational, models of a butterfly’s population dynamics can be constructed, each 3–7 year study was laborious and expensive. Moreover, in most cases the results were of limited usefulness for conservation, for they tended to describe the drivers of short-term fluctuations of a single, relatively stable, population, rather than the factors that determine the carrying capacities of different sites or which drive long-term population trends in species or assemblages across ranges. Most, however, provided the first insights into the key role played by the larval niche in butterfly population dynamics, owing to simple measurements made of the effect on larval survival of variation in two or more local growth-forms of the foodplant(s) or seral stage (e.g. Dempster 1969, 1983; Thomas 1974; Pollard 1979; Warren et al. 1986). Only in M. arion and Maculinea rebeli was variation in the quality of resource precisely quantified for every individual of the (1338, 1788 respectively) eggs whose survival was followed through each life-stage to adulthood (Thomas et al. 1998a, 2009). These latter studies enabled a series of mechanistic models to be constructed (reviewed by Thomas et al. 1998a; Clarke et al. 2005) in which habitat quality was a major variable, ranging from phenomenological or deterministic density-dependent interaction models (Hochberg et al. 1992; Mouquet et al. 2005a, b; Thomas et al. 2009) to others involving spatially explicit individual-based stochastic processes (e.g. Hochberg et al. 1994; Clarke et al. 1997).

With the realisation that variation in the quality of the larval resource was the known (or likely) driver of major population changes in studied species (Thomas 1984a), the focus of conservation research shifted to the less onerous precise description of species’ realised larval niches, and to examining the extent to which variation in the quality or abundance of these and the adults’ resources correlated with site occupancy or population size across a large number of habitat patches (for review of early studies see Thomas 1991). These studies also suggested that typically 10–40% of patches containing apparently suitable source habitat were not occupied by populations of colonial species (sensu Thomas 1984a) at any one time, and the constraint of low adult dispersal across landscapes was recognised early on (e.g. Ehrlich 1961; Thomas 1974, 1975, 1983a, b; Dempster et al. 1976; Dempster 1991; Warren 1987a; Thomas et al. 1992). Soon, a new science of meta-population ecology was developed by Hanksi and co-workers, largely using butterflies as model systems (for reviews see Hanksi 1999; Ehrlich and Hanski 2004). This also spawned numerous landscape-scale studies useful to practical butterfly conservation.

During the past decade, much research on butterflies has focused on adult resources, physiology, behaviour, and larval interactions with ants (myrmecophily) and parasitoids, providing an unparalleled knowledge of the multiple components in the environment, including the resource base exploited by most stages (Dennis and Van Dyck 2003, 2006; Turlure et al. 2010), that impact on a butterfly species’ populations and with which it interacts (see Settele et al. 2009; Dennis 2010). An important development has been to study and predict the impacts of climate change on the definition and availability of species’ realised niches, ranges and abundance (e.g. Settele et al. 2008), and the extent to which species are regionally adapted to their local environment, including the rate at which phenotypes can adapt to habitat availability, climate or other changes following strong regional selection, for example for increased dispersal (e.g. Hill et al. 1999; CD Thomas et al. 2001; Hughes et al. 2003).

Here we attempt to analyse which factors have caused butterfly populations to decline to extinction or increase in density by the one to two orders of magnitude often observed in the wild. Except in the context of landscape-scale (meta-population) studies, we restrict analyses to the many semi-natural and other sites studied where a species’ resources still apparently exist, rather than tabulate the fundamental destruction of biotopes (sensu Thomas 1991), such as the drainage and agricultural improvement of wetlands or the clearing of rain forests, that remains the main driver of butterfly extinctions and conservation priority in less developed regions of the world (e.g. Brown and Brown 1992; Stewart et al. 2007). In particular, we enquire whether just one or two key factors exist for study species that have such an over-riding impact on its population dynamics that all other interactions, while undeniably interesting, are of little relevance when it comes to manipulating habitats for practical conservation. In other words, we examine the evidence-based route successfully taken in entomological pest control, where the manipulation of one or two parameters may transform a site’s carrying capacity and hence a species’ abundance.

Intra-specific variation in population density

To understand the magnitude of variation that exists between the size of populations supported by a species’ optimum and sub-optimal (usually occupied) source sites, we here extend a preliminary (Thomas 1984a) analysis of the UK Butterfly Monitoring Scheme (BMS) time-series of species’ indexes of abundance, using the additional 24-years’ data now available at http://www.ukbms.org (Fig. 1). For each species, a comparison was made of the mean adult counts per 100 m of 5 m wide transect over 10–33 years on the five sites that consistently supported the highest density of adults, and the five sites that supported the lowest densities. To avoid artefacts or sampling bias, considerable care was taken to compare like with like. Thus comparisons were restricted to sites containing single populations and similar habitat types, for example to more-or-less homogenous grassland containing the larval foodplant in the cases of Melanargia galathea, Coenonympha pamphilus and Maniola jurtina, thus excluding woodland sites that contained occasional rides and glades; to calcareous grasslands for M. galathea and acid ones for Hipparchia semele; to sites where the transect was considered to sample the population of each species to an equal extent; and so on. In addition, for a site to qualify as supporting a low-density population, transect counts must record adults for at least nine out of ten consecutive years (see Thomas et al. 1994 for full criteria).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig1_HTML.gif
Fig. 1

Intraspecific variation in the mean densities of populations of eight butterfly species recorded on UK Butterfly Monitoring Scheme transects. Columns represent the mean population density (±SE) per 100 m of transect on the five highest-density sites and the five lowest-density sites, with the mean of the low density values = 1 (+SE). Methods follow Thomas (1991) and Thomas et al. (1994), with analyses restricted to sites occupied for 9 of 10 successive years by a species, and to those containing similar types of habitat, including larval foodplant, for each species in all sub-areas sampled by the transect. aMelanargia galatheabHipparchia semelecArgynnis paphiadPolyommatus coridoneManiola jurtinafCoenonympha pamphilusgBoloria euphrosynehPlebejus argus

The results, presented here for eight species (Fig. 1), confirm the tentative earlier conclusion that the density of adults in the largest populations of a butterfly species is consistently about a hundred times higher than the density in its smallest ones. The same is true of Maculinea nausithous, M. rebeli, Maculinea teleius, M. arion, Maculinea alcon, Melitaea cinxia, M. athalia, P. bellargus, Polyommatus coridon, Cupido minimus, Thecla betulae and H. comma where population estimates have been based on egg or larval counts and contained low (usually 0–15%) standard errors (Thomas 1974, 1983a, b, 1984b; Elmes et al. 1996; Thomas et al. 1986, 2009; Thomas and Lewington 2010; Warren 1987b, 1991). Comparing the single highest density BMS site with the single lowest density one typically showed a 1,000-fold difference.

Hereafter we assume that the observed two-order-of-magnitude difference in butterfly density between the largest and smallest populations of a species in a landscape reflects the spectrum of environmental quality that exists between its optimum sites and those of the lowest source quality that are nevertheless capable of supporting a predominantly closed population for at least 9–32 generations. There is abundant evidence (Dennis 2010) that some differences are explained by fixed, typically abiotic, attributes, such as aspect, soil depth, shelter and local climate that are impossible or very expensive for conservationists to manipulate. For example, due to variation in soil attributes, the two highest quality UK restoration sites for M. arion have consistently supported 42 times more adult butterflies per square metre of Thymus/Myrmica sabuleti grassland than the two poorest ones, even when the latter were optimally managed for 10–21 years (our calculation from Thomas et al. 2009). On the other hand, there are many hundreds of examples of apparently suitable sites where the population size of a non-migratory species (including M. arion) dramatically decreased or (less often) increased by two orders of magnitude over a short number of generations before stabilising at a new density, whilst neighbouring populations remained relatively stable (e.g. Thomas 1983a, b; Thomas et al. 1986, 2009; Warren 1987b, 1991). We illustrate one of many examples from a wide range of species and populations to be found in the UK BMS dataset, involving a hitherto unreported comparison of changes on two UK sites supporting P. coridon. As is typical of many such shifts (Thomas 1991), the larval foodplant Hippocrepis comosa remained abundant and apparently unchanged in density on both sites during this period (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig2_HTML.gif
Fig. 2

Examples of change by two orders of magnitude in the population density of Polyommatus coridon on single UK sites following: a gradual habitat degradation and b conservation management. Grey symbols, dotted line national trend for P. coridon; black lines and symbols population at a Swanage and b Winchester following conservation management (arrow). Dashed lines gap of 2–3 years in recording

Our premise for the discussion below is that the differences shown in Figs. 1 and 2 define the order of magnitude that a population parameter must explain if it can be considered a key factor that drives long-term population change in butterfly species. For it is in the major drivers that can convert a potentially high-quality site into being a sub-optimal (or extinct) one, and especially vice versa, that the practical conservationist should be interested. We consider candidate factors.

Identifying factors determining abundance and drivers of population change

For most studied species of butterfly, analysis of the factors regulating population size or the drivers of population change is not straightforward, since typically only one or a few candidate factors were measured per study. However, it is possible to make a crude meta-analysis of 30 European species that have been studied more completely (Table 2), involving the compilation of life-tables, or comparatively complete comparisons being made both of site attributes and populations, or where population size was altered following manipulation of a site’s environment through targeted conservation management or, in the case of Pieris rapae (Dempster 1969), pest control. For each species, we assigned a value of 0–3 to any of 11 non-exclusive variables that have historically been considered as a key population factor for butterflies, based on authors’ assessments of the importance of each.
Table 2

The principal source material for meta-analysis of Fig. 3

Species

Adult parameters

Larval parameters

Type of study

Tested in field?

Source

Nectar honeydew

Mating

Natality weather

Dispersal

Collectors

Shelter

Food density

Subset of quality food

Parasitoids

Site area

Pesticide

Carterocephalus palaemon

X

X

C

X

14, 20, 30–33,

Thymelicus action

a

a

C

s

11, 12, 14, 20, 18, 58

Hesperia comma

a

a

a

X

a

C

lsc

12, 14, 15, 25, 52, 53

Pyrgus malvae

a

X

X

X

C

X

35

Papilio machaon

X

a

a

Ilt

s

5, 6, 12, 14, 58

Leptidea sinapis

X

a

Ilt

lsc

12, 14, 40, 48, 58

Pieris rapae

a

a

X

Lt

X

1, 2, 3, 5, 12

Thecla betulae

a

Ilt

s

8, 12, 14, 20, 59

Satyrium pruni

a

Ilt

lsc

8, 9, 12, 14, 20, 58–59

Lycaena phlaeas

a

a

a

a

a

a

X

Ilt

X

4, 5, 14

L. dispar

a

X

a

X

Ilt

s

5, 7, 12, 14

Plebejus argus

a

a

a

X

C

s

20, 23, 24, 26, 57

Aricia agestis

a

a

X

X

C

X

28

Polyommatus bellargus

a

C

lsc

11, 12, 14, 18, 20, 37, 58

Maculinea arion

Ilt

lsc

5, 12, 14, 17, 19, 38, 58

M. rebeli

a

X

Ilt

X

17, 29

M. nausithous

a

a

Ilt

s

13, 14, 17, 49, 50

M. teleius

a

a

Ilt

s

13, 14, 17, 49, 50

M. alcon

a

a

a

X

Ilt

X

17, 39

Limenitis camilla

a

a

a

X

a

a

Ilt

X

12, 14, 21

Boloria selene

a

X

a

C

lsc

14, 16, 20, 27, 45, 58

B. euphrosyne

a

X

a

C

lsc

14, 16, 20, 27, 45

Agrynnis adippe

a

a

a

X

a

C

lsc

14, 16, 20, 27, 45, 47, 58

A. aglaja

a

a

a

X

X

C

X

14, 16, 20, 27

A. paphia

a

a

X

a

C

s

14, 16, 20, 45

Euphydryas aurinia

a

X

Lt

lsc

20, 22, 34, 51, 55, 56

Melitaea cinxia

a

X

Lt

X

14, 18, 20, 46,

M. athalia

A

X

Lt

lsc

12, 14, 41–44, 58

Hipparchia semele

a

X

X

C

lsc

20, 27, 58

Coenonympha tullia

a

X

X

C

X

36, 57

Columns 2–12, main variables assessed in population studies (✓ = precisely or crudely measured, a = anecdotal assessment, X = not measured), nectar = impact of available nectar on longevity or natality; column 13 type of study (Ilt = individual-based life-tables, Lt = life tables, C = correlations); column 14 key factor(s) validated or identified by testing in the field through conservation or other management at scale of landscape (✓lsc) or single sites (✓s); column 15 main sources by number: (1–4) Dempster 1967, 1968, 1969, 1971; (5) Dempster and Pollard 1981; (6) Dempster et al. 1976; (7) Duffey 1968; (8–14) Thomas 1974, 1975, 1983a, b, 1984a, b, 1991; (15–19) Thomas et al. 1986, 1996, 1998a, 2009, JA Thomas et al. 2001; (20) JA Thomas and DJ Simcox unpublished; (21) Pollard 1979; (22) Porter 1981; (23, 24) Thomas 1985a, b; (25) Thomas and Jones 1993; (26) CD Thomas et al. 1999; (27) Elmes and Thomas 1992; (28) Bourn and Thomas 1992; (29) Hochberg et al. 1994; (30–33) Ravenscroft 1994ac, 1995; (34) Lewis and Hurford 1997; (35) Brereton 1997; (36) Dennis and Eales 1999; (37) Roy and Thomas 2003; (38, 39) Mouquet et al. 2005a, b; (40–47) Warren 1985, 1987ac, 1991, 1992, 1994, 1995; (48) Warren et al. 1986; (49, 50) Wynhoff 1998, 2001; (51) Wahlberg et al. 2002; (52) Wilson et al. 2009; (53) Davies et al. 2005; (54) Revels 2006; (55) Bulman et al. 2007; (56) Klapwijk 2008; (57) Dennis 2010 and references therein, (58) Thomas and Lewington 2010, (59) Bickmore and Thomas 2000)

The results (Fig. 3) support previous reviews (e.g. Thomas 1991; New et al. 1995), as well as Hanski’s (1999) statement that the realization that the constraints on species imposed by variation in the quality of the larval habitat and the ability of adults to colonise new patches across landscapes were “the two major reorientations in butterfly biology and conservation in 20 years”. Below, we consider these two and certain other putative drivers at a detail commensurate with their ranking in Fig 3. Thus the potential impacts of butterfly collectors and insecticides, which both scored zero and have long been dismissed as negligible drivers of change (Thomas 1984a), are not discussed. They were, however, included in the meta-analysis because, for >100 years and a decade respectively prior to 1970, they were the main factors believed to threaten butterfly populations, and mitigating their impacts formed the rationale for many unsuccessful conservation projects.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig3_HTML.gif
Fig. 3

The main drivers of long-term population trends in 30 European butterfly species where full population dynamic or similarly intensive studies were made and/or targeted conservation or pest control management has succeeded. Scores used in meta-analysis: 3 key driver, 2 secondary driver, 1 detectable effect, 0 no effect detected. See Table 2 for sources

Weather, climate change and population size

The observed and predicted impacts of climate change on different species are reviewed elsewhere by Dennis (1993), Parmesan et al. (1999), Roy et al. (2001), Warren et al. (2001), Thomas et al. (2006) and Settele et al. (2008), and are not repeated here. However, variation in the attributes and breadth of a species’ larval niche under different climates is briefly considered below, and it is relevant to note that the short-term fluctuations (as opposed to long-term trends) of monitored species, which tend to be synchronised across regions (Pollard 1991) and are widely accepted as being driven mainly by variation in weather (Pollard and Yates 1993; Roy et al. 2001), are typically two orders of magnitude smaller than the consistent differences or long-term trends in population density between sites reported here (Figs. 1 and 2). Thus although short-term fluctuations are of significantly greater amplitude towards species’ range margins (Thomas et al. 1994), the mean temporal Coefficient of Variation (CV) on the stable sites of the species represented in Fig. 1 ranges from 0.62 (M. jurtina) to 0.91 (P. coridon) (see Thomas et al. 1998a for CVs of all UK species that meet criteria for analysis).

Shelter

A lack of shelter scores as a depressor of population size in many studies. With the exception of Dover’s (e.g. Dover 1996; Dover et al. 1997) studies in open farmland, this is the least precisely defined of the 11 variables considered, and with increased knowledge each example could probably be reclassified as equating to ‘within-site larval quality’ or ‘low adult dispersal’, or both. For variation in shelter can affect butterfly populations in either the immature or adult stages: in the former by altering the microclimate, especially temperature, of the larval habitat; in the latter, by affecting the propensity of adults to fly across exposed or more sheltered examples of the matrix that lies between habitat patches, or within sites by providing shelter from inclement weather or refugia at night (Dover et al. 1997; Dennis 2010). Although often ill-defined (except by Dover 1996), shelter is a useful parameter to identify for the practical conservationist, for it can be manipulated on many sites, for example through scrub or woodland management (e.g. Thomas 1975; Dover et al. 1997).

Population-scale factors

Interactions with enemies or resources

Very few populations of recorded butterflies (Asher et al. 2001) exhibit the major cycles in abundance or patterns of short- to medium-term boom-bust changes that are characteristic of reciprocal interactions with resources or enemies in certain other insect taxa or in simpler communities breeding on intensively cultivated land. Moreover, most short-term fluctuations are attributable to weather effects (above). A striking exception, postulated by Pollard and Yates (1993) and confirmed empirically by Revels (2006), is the 4–6 year population cycling of Celastrina argiolus and its specific ichneumonid parasitoid Listrodomus nycthemerus (Fig. 4). Even here, it is possible that the equilibrium level around which both insects cycle is determined by the density or quality of the host’s habitat. Similar dynamic interactions have been claimed for certain Melitaeninae or related fritillaries, but at present the evidence is equivocal (e.g. Porter 1981; Klapwijk 2008). In other butterfly species, cycles or declines due to over-exploitation of a resource, for example the weak cycling just detectable in M. arion populations around each site’s equilibrium level (Thomas et al. 2009), are orders of magnitude below that needed to explain observed major population changes.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig4_HTML.gif
Fig. 4

Four to six year population cycles of Celastrina argiolus, tracked and apparently driven by its specific parasitoid Listrodomus nycthemerus. Data from Revels (2006), plotted here as 3-generation running averages to smooth seasonal fluctuations of spring and summer generations

Adult resources

Limited or declining adult resources are frequently evoked as the cause of small populations or long-term declines, but apart from shelter, the evidence for this is weak. Indeed, most putative factors (nectar, mating and thermoregulatory spots) scored zero in the meta-analysis of 30 species studied in the field (Fig. 3). Moreover, their impact, in the relatively few cases established under laboratory conditions, typically explains a <1 to 3-fold (under extreme unnatural conditions up to 7-fold) changes in the egg population (Boggs and Ross 1993; Fischer and Fiedler 2001; Bergstrom and Wiklund 2002; Mevi-Schütz and Erhardt 2003; Jervis and Boggs 2005) rather than the 50–100-fold or greater variations (Figs. 1, 2) which we seek to implicate here. Thus intra-specific variation in mean female natality seldom varied by more than 2- or 3-fold between years and sites in species for whom life-tables have been compiled (Table 2), and when it did, the effect was often diminished before the next adult emergence by density-dependent mortalities (or enhanced survival) in the intervening larval stage (e.g. Dempster 1983; Pollard 1991; Nowicki et al. 2009; Thomas et al. 2009). Moreover, observed variation in oviposition and population numbers is strongly correlated with the weather (e.g. Gossard and Jones 1977; Thomas 1974; Pollard 1979, 1991; Roy et al. 2001; Thomas et al. 2009) or with early adult death by enemies (Dempster 1983), both of which contribute to the mean 4–7 day survival rate of non-hibernating adults in temperate regions measured in the field, compared to the 20–30 days when protected in captivity. Nevertheless, two adult resources that are still influential in conservation planning are briefly considered.

Nectar. Prior to the 1970 s, a lack of nectar was hypothesised to be the third major factor (after collectors and insecticides) driving declines in the UK, to an extent that senior conservation advisors proposed the planting of (exotic) buddleja and other nectar sources as the main strategy to improve nature reserves where butterflies were declining. To our knowledge there is no scientific study, in which the mechanisms causing population dynamic change in a butterfly have been measured, that implicates variation in nectar as a driver. Certainly, several studies have reported a positive correlation between population size and the abundance of nectar sources on study sites (see Dennis 2010). On the other hand, no fewer than 12 of the 30 species included in the meta-analysis (Fig. 3 and references therein) show a significant inverse correlation between population size (or change) and the availability of their preferred adult nectar sources. This is perhaps simply explained by the fact that the optimum habitat for oviposition and larvae of these latter species involves growth-forms of the larval foodplant (or ant resource) existing either in very early (H. comma, Plebejus argus, P. bellargus, M. arion, M. rebeli, Boloria euphrosyne, Argynnis adippe, M. cinxia, H. semele) or late (Thymelicus acteon, M. nausithous, Argynnis paphia) seral stages of their biotopes, in which flowers generally occur at low densities: in contrast, nectar is usually most abundant in mid-successional swards (and other biotopes) (Erhardt 1985), coinciding with the optimum larval niches of Pyrgus malvae, M. teleius, M. alcon, Boloria selene, Agrynnis aglaja, Euphydryas aurinia and Coenonympha tullia of the species in Fig. 3. It should be noted that most of the above correlations were obtained from studies made in the UK or the northern half of Europe: several species that occupy flower-poor early seral stages in northern Europe occupy cooler mid-successional stages with abundant flowers under warmer climates further south (Thomas et al. 1998b; JA Thomas et al. 1999).

Mate-location. To our knowledge, there is no evidence of a direct impact on population size in European butterfly species from the availability of preferred mating sites. Moreover, although the females of certain species may be reluctant to lay eggs when not mated (Karlsson and Van Dyck 2009), the incidence of sterility is generally less than 2% of the egg population in even the smallest studied populations, and as described above, variation in natality, for which there is little evidence of any effect from the nutritional value of male nuptial gifts (Bergstrom and Wiklund 2002), or the reluctance of unfertilised females to oviposit, explains no more than small-scale fluctuations in well-studied species. For the same reason, the possibility of increased emigration of adults from patches with few mating sites (or other adult resources) does not explain variation in population size in species for which life-tables have been constructed. Nor would it necessarily be a bad thing, at the scale of meta-populations (below), if this were a major variable.

Larval habitat quality

In all but one (i.e. 97%) of the well-studied species in Fig. 3, and in the large majority of other species where the ecology of the young stages has been studied, the female restricts or concentrates oviposition on a subset of the larval foodplant that is growing in definable and predictable situations or growth-forms (see Thomas 1991, 2007; Dennis 2010; Thomas and Lewington 2010). The attributes of the preferred subset vary greatly between species. For example, each of the five European Maculinea species restricts oviposition to initial foodplants that possess flowerbuds of a narrow size range, development stage and prominence (Thomas and Elmes 2001); Carterocephalus palaemon oviposits on a minority of Brachypodium plants whose leaves are not only tall and rich in nitrogen, but also growing in warm sheltered spots (Ravenscroft 1994b); Aglais urticae (Pullin 1987) and Polyommatus icarus (Dennis 1985) use young, nitrogen rich growths of their foodplant, whereas Aricia agestis selects plants with exceptionally succulent leaves (Bourn and Thomas 1992); T. acteon selects only the largest tussocks of Brachypodium pinnatum, and of these only clumps growing in warm sheltered spots (Thomas 1983b); B. selene exploits medium-sized Viola plants growing in moist soil whereas B. euphrosyne selects younger, smaller-leaved Viola growing in warmer spots (Thomas et al. 1986); in woodland, the larval niche of many species whose young stages inhabit the ground or shrub layer is restricted to examples of foodplant growing within narrow limits of light or shade (e.g. Pollard 1979; Shreeve 1986; Warren 1987b,c, 1995; Thomas 1991; Thomas et al. 1986). And towards the northern edges of their ranges, a great many species oviposit only on examples of foodplant that are growing in exceptionally warm microclimates, typically restricting the larval stage to slopes with warm aspects and within them to spots of early successional vegetation, such as very short or sparse grassland swards or fresh clearings in woodland (Thomas 1991, 1993; Thomas and Lewington 2010).

Where measured, the optimum form of the larval resource is closely correlated with enhanced larval survival or fitness in the wild (e.g. Thomas et al. 1998a). So restrictive is this specialisation that, in at least 20 of the species included in the meta-analysis (Fig. 3), no correlation was found between the absolute abundance of larval foodplant and butterfly population size across sites or over time, and in the cases of H. comma, where oviposition was restricted to small Festuca ovina plants surrounded by bare ground, and in M. nausithous which oviposits in late successional wetlands where Sanguisorba officinale grows tall but typically sparsely, a significant inverse correlation was found between foodplant and butterfly densities (Thomas 1984b; Thomas et al. 1986). On the other hand, in all 20 species used in the meta-analysis where the availability of the preferred larval niche was adequately quantified,1 a strong correlation existed between this, often minority, subset of the larval resource and butterfly population size. Importantly for conservation, the association in all these species explained a spectrum of populations sizes that spanned two or more orders of magnitude (Figs. 1 and 2), both between sites or within individual patches where management changed over time. In the few species where the preference of ovipositing females and larvae was precisely quantified across the full spectrum of its resource, ranging from optimum via sub-optimal sources to sink habitats, the correlation between larval habitat quality and butterfly density was close, with values of R2 = 51% (P. bellargus), 54% (M. cinxia), 62% (T. acteon) and 99% (M. arion) (JA Thomas et al. 2001; Thomas et al. 2009).

The importance for conservation of identifying and manipulating this specialisation is illustrated in two case histories. In northern parts of its range, larvae of the bivoltine butterfly P. bellargus are restricted during the autumn/spring-feeding generation to a (usually small) subset of H. comosa plants growing in sheltered south-facing depressions where the sward is also very short (Fig 5a). These represent the warmest hotspots available within sites (Thomas 1983a), and the relative sizes of P. bellargus populations in the UK are closely correlated, across two orders of magnitude, with the number of foodplants growing in optimum or sub-optimal growth-forms (JA Thomas et al. 2001). In May–June, the resultant adults oviposit on plants growing in the same narrow range of absolute temperatures to those selected by autumn-flying adults, but because the weather is warmer, the eggs are distributed over a wider, partly different subset of foodplants, including many growing in relatively cool and unsheltered spots (Fig. 5b). On an average UK site, this permits the summer-feeding generation of larvae to exploit roughly twice as many H. comosa plants than in autumn-winter, and may explain why the mean adult population size of P. bellargus was about twice as numerous in the second generation on monitored sites in 1976–2002 (Roy and Thomas 2003).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig5_HTML.gif
Fig. 5

Seasonal shifts in the larval niche of the bivoltine thermophilous butterfly Polyommatus bellargus in the UK, on a site where the distribution of its perennial foodplant Hippocrepis comosa was constant. a Autumn–spring-feeding larval generation, b summer-feeding larval generation; from Roy and Thomas (2003) where categories of shelter are defined (3 most sheltered), grey columns growth-form of Hippocrepis exploited in autumn–spring only, white growth-forms used by both generations, black summer only. c Mean density (± SE) of P. bellargus in each adult generation on UK sites sampled by the UK butterfly Monitoring Scheme in 1976–2001; dashed lines predicted carrying capacity of sites in each generation calculated from the proportion of Hippocrepis plants available for exploitation by the preceding cohort of larvae

Over much of northern Europe, P. bellargus populations declined and experienced frequent local extinctions during the first 80 years of the twentieth century, when many farmers abandoned unfertilised south-facing slopes and, after the mid 1950 s, rabbits disappeared as an effective grazing force due to myxomatosis. In the UK, the rate of population extinctions was so steep that by 1980 an extrapolation suggested the national extinction of this species early in the current century (Thomas 1983a). In practice, recognition that populations of P. bellargus survived only on sites where H. comosa grew in close-cropped hotspots, and that large populations persisted only where many plants occurred in that condition, caused wardens and sympathetic farmers to increase grazing on declining and former sites, applying regimes which conservationists had shunned because they reduced the flowering of most plant species (the more recent discovery that P. bellargus will inhabit taller turf during summer negated that conflict). This restored grazing, coupled with the gradual return of rabbits, transformed the sward structure of calcareous grasslands in favour of this butterfly’s larval niche, as evidenced by the results of repeat surveys made of all potential sites across 80% of P. bellargus’ occupied UK range in 1978 and 1999 (Thomas 1983a; JA Thomas et al. 2001). Coinciding with this shift (and incidental reduction in nectar availability), many of the ca. 75 surviving populations of 1980 increased greatly in numbers and spread to found about 75 new colonies by 1990 and a further 100 populations a decade later, mainly colonising former sites (Fox et al. 2006). A similar 4-fold recovery in site occupancy occurred over the same period with H. comma in the UK, for the same reasons (Davies et al. 2005), aided also, in both species, by a warming climate which broadened the larval niche and hence the number of exploitable foodplants (JA Thomas et al. 1999).

The evidence that the recovery of P. bellargus and H. comma in the UK resulted initially through targeted conservation management to restore their larval niches in grassland is persuasive, yet based on correlations. Greater certainty exists about the restoration of M. arion to the UK, following national extinction in 1979, for new grazing regimes were based on precise niche and population models resulting from an intensive 6-year study of the mechanisms affecting natality and survival in all stages of the life-cycle (see Introduction). Again, that study indicated that this most specialised of butterflies was an even greater specialist during its larval period than had been appreciated, and that apparently minor variation to the structure of its grassland habitat altered both the intrinsic growth rate of a population (λ) and the carrying capacity (K) of its sites from <1 to 5.9 and from 0 to 90,000 eggs per hectare respectively (Thomas et al. 2009), with both parameters being closely correlated (R2 = 85%) on the 13 sites where both have been measured (Fig. 6). Although confined to different soils—and requiring the additional resource of the ant M. sabuleti to coexist with the initial larval foodplants (Thymus or Origanum vulgare), occurring for optimal survival as high density, small colonies containing large-bodied workers and few or no queens—the vegetation structure required by M. arion larvae resembles that of P. bellargus and H. comma at northern latitudes and high altitudes, and requires similar grazing management (Thomas et al. 2009). Although low survival was recorded with other species of Myrmica, early models predicted that a threshold of 68% of the larval population must be adopted by M. sabuleti for the butterfly’s intrinsic rate of increase to exceed one. M. sabuleti is a thermophilous ant that requires warm short open swards under UK climates (Fig. 7). For the same reasons that P. bellargus and H. comma declined across northern Europe, so did M. arion’s specific ant host, to the extent that it was undetectable on the large majority of former sites when surveyed in the early 1970 s. Although research results came too late to save the last small UK M. arion population from extinction, attempts to restore its optimum larval habitat, through grazing sites to < 2 cm tall during the critical spring and autumn months for M. sabuleti larval development, began in the mid seventies and are currently being applied to about 100 former or neighbouring sites. The return of M. sabuleti over a 35 year period is illustrated for one restoration site (Fig. 7), and is typical of all others (Thomas et al. 2009): it is pertinent to note that the niche of M. sabuleti has broadened under the warmer climates since the early 1990 s (Fig. 7, dashed curve), and today the minimum density required to support M. arion encompasses sward heights of up to 4 cm tall, easing the delivery of successful conservation management.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig6_HTML.gif
Fig. 6

Relationship between values of the observed intrinsic growth rate (λ) and the observed carrying capacity (or equilibrium level) (K) of Maculinea arion populations following colonisation of 13 unoccupied conservation sites in the UK. Values of K represent eggs per m2, based on the mean of the three highest values of populations that stabilised and fluctuated round an equilibrium level (CV < 0.5, see Fig. 8). Plotted from data in Supporting Online Material (Thomas et al. 2009) and D.J. Simcox (unpublished): obs K = −2.17 + 1.67 (obs λ), R2 = 85%; P = 0.000 (omitting outlier, R2 = 59%; P = 0.004)

https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig7_HTML.gif
Fig. 7

Changes in the niche and abundance of the ant Myrmica sabuleti under past (solid curve 1974–1989) and current warmer (dashed curve 1990–2008) UK climates on one typical Maculinea arion restoration site. Black dots years when site did not support M. arion; grey dots years supporting M. arion; triangles ungrazed controls; horizontal dashed line minimum predicted density for M. arion to have a positive intrinsic growth rate (λ > 1). From Thomas et al. (2009)

By the early 1980s, a simple mechanistic model (Thomas et al. 2009), in which four of the five variables described different aspects of larval survival in ant nests (proportion of larvae adopted by M. sabuleti, other or no Myrmica; weight of ant brood; density-dependent mortalities through starvation in ant nests; impact of drought on ant carrying capacity) was tested by introducing Swedish M. arion to Dartmoor (Site X), where the original population study had been made and where the predicted carrying capacity (and value of λ) had been raised through the impact of altered management on M. sabuleti populations by nearly 10-fold. Figure 8 shows that for the 22 years over which this new population was monitored, the actual number of M. arion closely followed the long-term theoretical predictions. Further validation was obtained from the second longest-running population, introduced to Green Down in 1992, where the predicted values of λ and K for M. arion were 0.3 and 0 prior to conservation management (representing sink habitat), but quickly rose to >3 and nearly 4,000 eggs per hectare, respectively, under the prescribed grazing. Again, the introduced population of M. arion closely followed long-term predictions derived from the quality of its larval habitat. The same was true of model predictions of 109 annual population changes on other sites in 1983–2009 (N = 109 population changes on 21 sites; GLM adjusted R2 = 77.7%, P = 0.000; from Thomas et al. 2009 plus 10 additional datapoints and two new sites for 2008–2009 changes). At the time of writing, M. arion has spread to >30 prescriptively managed sites in the UK, two of which have supported the highest known densities in the world of this globally threatened species for the past decade (Thomas et al. 2009). Together, these datasets establish that variation in the occurrence of the larval niche (and the occasional impacts of drought on larval resources) is the overriding variable that determines long-term trends and average population size in this Red Data Book species.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig8_HTML.gif
Fig. 8

Predicted (dashed line) and observed (solid line) population changes of Maculinea arion on two UK conservation sites. Broken dotted lines change in estimated carrying capacity of sites before and after conservation management (KGreen Down = 0 prior to 1990). Solid symbols Site X, Dartmoor: 1972–1978 = the last native UK population during the period of intensive population study; 1983– = introduced population from Sweden. Open symbols Green Down, Somerset. Predictions of population changes based on a mechanistic model describing annual variation in the quality of the resource, including impacts of drought in 5 years, with predictions made sequentially from dates of introductions (1983, 1992) without correction for observed butterfly numbers in subsequent years; data and model from Thomas et al. (2009) plus 2009 data for Green Down

There is similar evidence of conservation success among other butterflies whose larval niche was precisely described and then deliberately created or increased through the management of individual sites. The longest-term or best documented examples of populations increasing by 10- to >100-fold in numbers after targeted management again come from the UK or the Netherlands. They include in woodland, scrub or scrub-grassland: Satyrium pruni (Thomas and Lewington 2010), T. betulae (Bickmore and Thomas 2000), Hamearis lucina (MS Warren pers comm), Leptidea sinapis (S Jeffcoate pers comm), M. athalia (Warren 1991) and four species of Viola-feeding fritillary (Elmes and Thomas 1992; Asher et al. 2001; Thomas and Lewington 2010; DJ Simcox unpublished); in more open grasslands P. coridon, H. semele (Elmes and Thomas 1992) and E. aurinia (Bulman 2007); and in wetlands M. nausithous and M. teleius (Wynhoff 1998, 2001), Lycaena dispar (C Van Sway pers comm), and Papilio machaon (Fox et al. 2006). Only in S. pruni, M. athalia, L. dispar and perhaps P. coridon has this management been applied to enough sites to reverse or halt the national decline, but these examples provide evidence of the importance of maintaining this component of a species’ habitat.

Although every species studied in this way possessed a unique, species-specific fundamental niche (albeit one that may vary under different climates), among the more hopeful results of recent conservation projects is that the targeted restoration of one declining species’ habitat has often resulted in an improved, if not optimal, environment for other specialists, that had been declining for the same general reasons of altered land-use. In other words, where they belong to a similar guild, it has proved possible to manage for declining assemblages of butterflies (and other wildlife). Most accounts are anecdotal, but quantified examples include 1-2 order of magnitude increases by seven other Biodiversity Action Plan (BAP) butterfly species (B. euphrosyne, B. selene, A. adippe, H. semele, P. malvae, Erynnis tages, T. betulae) on UK sites managed for M. arion.

A more rigorous test of understanding a species’ requirements is to attempt to create them from scratch on intensive arable farmland, or on the raw bedrock or ballast of major constructions, such as on the verges, embankments and cuttings of new roads or railways. It is beyond the scope of this paper to describe examples in detail, but most successful ones have involved inclusion into the design of the precise larval niches of an endangered butterfly species or, more often, of an assemblage of scarce or threatened species, including their preferred topographies, soil depths and nutrient levels (e.g. Morris et al. 1994). Successful UK examples include the restoration of species-rich calcareous grassland to the arable fields of Twyford Down and the adjoining creation on chalk ballast of >2 km grassland on the former A303 near Winchester, UK following the construction of the M3 in the 1990s, creating new sites that today support high densities of BAP species P. coridon, C. minimus and E. tages (Snazell et al. 2001; Thomas 2001; Thomas and Lewington 2010); the creation of scrub and clay grassland habitats on former arable fields isolated by the M40 in Oxfordshire, which today support high densities of T. betulae and the UK’s largest population of S. pruni (Bickmore and Thomas 2000 and unpublished 2007–2009 monitoring); and the creation of two sites on new UK railway constructions that today support medium-sized populations of M. arion and other BAP species (D.J. Simcox, unpublished; Thomas and Lewington 2010).

Metapopulation-scale factors

Adult dispersal

A lack of occupancy of certain of a species’ apparently suitable habitat patches was recognised early in empirical studies of butterflies, and attributed to low adult dispersal (e.g. Ehrlich 1961). Convincing empirical evidence for low vagility comes from the rates of spread of certain species after colonisation or introduction to unoccupied landscapes containing multiple islands of the species’ larval habitat. Two early examples found that S. pruni and M. athalia spread at rates of about 1 and 1.6 km a decade (Fig. 9) in landscapes containing patches of variable habitat quality, the former being a similar rate to that recorded in P. argus after introduction to the Dulas valley, Wales, and the latter to the spreads of P. bellargus, H. comma and M. arion following the restoration of their habitats in patches across landscapes in England (Thomas 1991; Thomas et al. 1992; Davies et al. 2005; Thomas and Lewington 2010). Similarly, only one (Polygonia c-album) of the UK butterflies whose climate envelope limits have shifted >100 km northwards with warming climates in recent decades has fully tracked the availability of new habitat; in contrast, 34 out of 46 study species failed to expand at all (Warren et al. 2001; Hill et al. 2009).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig9_HTML.gif
Fig. 9

Observed colonisation rates of habitat patches of aSatyrium pruni, bMelitaea athalia following conservation management of sites across landscapes or introduction to new landscapes. From Thomas et al. (1992)

Field-based evidence supporting the predicted importance of a viable meta-population structure by stochastic patch occupancy (SPOM) and other models for species’ persistence in more typical landscapes has been elusive, but applies to at least nine examples listed by Schtickzelle and Baguette (2009). Most involve Melitaeninae and related fritillaries (e.g. Hanski 1994; Bulman et al. 2007), with examples also from the Lycaenidae and Hesperiidae (CD Thomas et al. 2001; Davies et al. 2005)

The remaining evidence-base for metapopulation viability is too well established (e.g. Hanski 1999) to need repeating. Here we return to the less familiar impact of variation of larval habitat quality within patches in a landscape, in this case in the context of meta-population dynamics.

Patch area, isolation and habitat quality

Theoretical and empirical meta-population studies have typically considered the size, number, distribution or dynamics of a species’ habitat patches in a landscape, and assumed that every source patch is of equal quality (Dennis 2010). We have seen, however, that the highest-quality sites consistently support a population density that may be 100 times greater than low quality ones of the same size (Figs. 1 and 2). On the few occasions during meta-population studies when within-patch variation in a species’ larval habitat quality was simultaneously measured precisely (as opposed to the simple recording of nectar sources, larval foodplant density or other assumed correlates with population λ or K), it was generally the single variable that explained the highest proportion of patch occupancy, with site isolation typically a close second. We illustrate the example of P. bellargus in the UK (Fig. 10): similar patterns are reported for C. tullia (Dennis and Eales 1997, 1999), Thymelicus acteon and M. cinxia (JA Thomas et al. 2001), Melitaea aurinia (Betzholtz et al. 2006) and M. alcon (Habel et al. 2001). JA Thomas et al. (2001), Dennis (2010) and Turlure et al. (2010) point out that in many meta-population studies, site area is incorrectly used as a surrogate for habitat quality: in field studies where both parameters were independently measured, site area explained few patterns of site occupancy (e.g. Fig. 10).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig10_HTML.gif
Fig. 10

Occupancy pattern by Polyommatus bellargus of all (95) sites containing Hippocrepis comosa in swards of varying structure in a 1960 km2 region of Dorset and Wiltshire, UK. The size, isolation and within-patch habitat quality for spring–autumn feeding larvae was measured on every site. From JA Thomas et al. (2001)

The high explanatory value of larval habitat quality rather than patch size in metapopulations is attributable to the dramatic potential difference in a species’ carrying capacity per unit area of site. For example, in the case of M. galathea (the species in Fig. 1 showing the greatest range of population densities), it would on average require a 586 ha patch of low quality source habitat to produce the same-sized population to that consistently supported by 1 ha of the species’ optimum habitat. The equivalent patch sizes of poor habitat for the other species depicted are 278 ha (H. semele), 139 ha (M. jurtina), 124 ha (P. coridon), 119 ha (C. pamphilus), 73 ha (Boloria euphrosyne), 56 ha (P. argus), and 39 ha (A. paphia). There are three mechanisms why high quality patches are more likely to be occupied and to contribute to a species’ persistence across landscapes, compared with low quality ones (JA Thomas et al. 2001):

(1) Probability of extinction. In theory and in practice, individual sites in a landscape that support small or low density populations have a higher probability of extinction (Hovestadt and Poethke 2006; Hovestadt et al. 2010). For example on BMS and other monitored UK sites, the 27 patches that initially supported a small (mean density 7.9 adults per km transect, max 17.2) population of B. euphrosyne experienced a combined total of 49 apparent extinctions (and 26 recolonisations) during the 10–28 years that they were monitored, with the average population occupying a site for just 5.9 (±SE 0.5) years. In contrast, the 13 sites that initially supported populations with B. euphrosyne indexes of >45 (max 61) adults per km transect experienced no extinctions in the 14–38 (mean 22) years during which they were monitored (our calculation from http://www.ukbms.org and unpublished data).

(2) Number of emigrants. All else being equal, a habitat patch that supports a hundred times higher densities of a butterfly species than a same-sized low-quality patch is likely to send two orders of magnitude more emigrants into the landscape, with a correspondingly higher likelihood that one or more individuals will reach a new patch. In reality, individuals are more likely to emigrate from very low density populations (due to rejection of low quality habitat or failure to encounter mates) or from very high density ones (due to density effects); site area and edge effects may also modify emigration (Hovestadt and Poethke 2006; Gros et al. 2009; Hovestadt et al. 2010).

(3) Probability of establishing a new population. Due to the close correlation between the intrinsic rate of increase (λ) and the carrying capacity (K) of habitat patches of varying quality for a species (Fig. 6), it is likely that an immigrant reaching a new or vacant habitat patch in a landscape has a higher probability of founding a population in a high quality (high K) habitat than in a low-grade source patch. This is because growth over the first few generations is faster in the former situation: for example, the observed initial growth rate of M. arion was 55 times greater on the highest quality patch compared with the lowest quality example of source habitat created on UK nature reserves (Fig. 6). Thus, in optimum habitat, a colonising population is likely to move quickly through the vulnerable period of very low numbers, when stochastic and inverse density-dependent factors make extinction likely. We can find little empirical evidence to support or reject this deduction, but it is suggested in theory by model simulations of immigrants’ success (Thomas and Hovestadt in prep), where habitat patch quality was parameterised to encompass the spectrum of values for λ measured on actual M. arion sites, and immigrant females were assumed to have mated, but thereafter experience the observed stochastic and inverse density-dependent mortality (e.g. in survival, natality, and unequal sex ratios) observed for butterflies at low densities (Thomas et al. 2009 and unpublished data). It is beyond the scope of this paper to provide full results, but a single typical simulation is presented in Fig. 11. In this example, the intrinsic rate of increase (λ) for low quality habitat (Fig. 11a) slightly exceeds the minimum observed for source habitat of M. arion in the field, yet it took 85 years to establish a population above a threshold size where persistence was likely, while after 100 years the population had reached only 20% of the carrying capacity for the site. In contrast, the same flow of immigrants invading mid-quality habitat (Fig. 11b, λ = 2.5) established a population within 22 years, and fluctuated around the site’s carrying capacity in years 30–100, even though the size of this site was, by definition, under half that of Fig. 11a. Other simulations indicate that the probability of establishing a new population increases only marginally across the upper half of the spectrum in patch quality observed in M. arion (Fig. 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9341-z/MediaObjects/10841_2010_9341_Fig11_HTML.gif
Fig. 11

Model simulations of the probability of founding a population in a patch of new habitat in a landscape. On average once every 10 years, an immigrant (fertilised, if ♀) reaches a patch with a carrying capacity (K) of 1,000 adults. a Low quality habitat (λ = 1.1), b medium–high quality habitat (λ = 2.5). From Thomas and Hovestadt (in prep)

Conclusion

In this analysis, variation in the intrinsic quality of habitat preferred by the larval stage of most of the better-studied species of temperate butterfly emerged as the most important factor determining the size and persistence of populations. Not only does the presence of optimum source habitat result in greatly enhanced intrinsic rates of population increase and carrying capacity within sites, but at the scale of a meta-population, there is field evidence that populations which occupy high quality patches are more persistent and release more emigrants into the landscape than is the case with those breeding in low quality source habitat; and in theory, new patches of optimum habitat are also more likely to be successfully colonised by immigrants.

Although the density-dependent and other mechanisms responsible for increases in λ and K are understood in very few species, it is sufficient for the practical conservationist to identify the preferred immature habitat of a valued species (which may vary across species’ ranges as well as under different regional climates) and, if possible, manipulate sites to encourage the generation or persistence of near-optimum conditions. Although in most species certain sites will be of intrinsically higher potential due to abiotic attributes that are expensive or impossible to change, the optimum habitat of most species studied correlates with a subset of their foodplants that is growing in a narrow seral stage in grassland, heathland, woodland or fenland successions. The latter properties can be manipulated by management, and there are several examples in most biotopes where targeted conservation management has resulted in the increase of a threatened species strongly against the national trend. Our studies of this variable at the meta-population scale suggest that conservationists would do well to spend their limited resources in improving the habitat of low- but potentially high-quality sites in landscapes, but that there are diminishing returns from striving to convert good sites into great ones. In practice, it appears sufficient to generate habitat in the upper half of the spectrum spanning the poorest to the highest quality habitat for a species or assemblage: across the lower half, there is a high probability of extinctions on occupied sites and a low chance that new sites will be colonised; beyond mid-quality habitat the chances of persistence and establishment are high and do not greatly improve for optimum conditions, although the latter does generate additional emigrants. On balance, on current (incomplete) knowledge, it is more effective to spend limited resources to achieve adequate, upper-range quality habitat on several sites for a species or assemblage in a landscape than to spend the same resources achieving the ultimate on a single site.

Footnotes
1

Thymelicus acteon, Hesperia comma, Leptidea sinapis, Thecla betulae, Satyrium pruni, Aricia agestis, Polyommatus bellargus, Maculinea arion, M. nausithous, M. teleius, Boloria selene, B. euphrosyne, Agrynnis adippe, A. aglaja, A. paphia, Euphydryas aurinia, Melitaea cinxia, M. athalia, Hipparchia semele, Coenonympha tullia. See Table 2 for references.

 

Acknowledgments

Research was conducted within the project CLIMIT (Settele and Kühn 2009), funded through the FP6 BiodivERsA Eranet by the German Federal Ministry of Education and Research (JAT), the French ANR (TH), and the UK NERC (DJS). We thank RHL Dennis for comments and suggestions.

Copyright information

© Springer Science+Business Media B.V. 2010