Journal of Insect Conservation

, Volume 15, Issue 1, pp 83–93

Recent trends in butterfly populations from north-east Spain and Andorra in the light of habitat and climate change

Authors

    • Butterfly Monitoring SchemeMuseu Granollers-Ciències Naturals
  • Ignasi Torre
    • Butterfly Monitoring SchemeMuseu Granollers-Ciències Naturals
  • Jordi Jubany
    • Butterfly Monitoring SchemeMuseu Granollers-Ciències Naturals
  • Ferran Páramo
    • Butterfly Monitoring SchemeMuseu Granollers-Ciències Naturals
Original Paper

DOI: 10.1007/s10841-010-9325-z

Cite this article as:
Stefanescu, C., Torre, I., Jubany, J. et al. J Insect Conserv (2011) 15: 83. doi:10.1007/s10841-010-9325-z

Abstract

Although butterfly declines have been reported across Europe, no assessment based on detailed quantitative data has ever been made for any extensive area in the Mediterranean Basin. In 1994, a Butterfly Monitoring Scheme was launched in Catalonia (NE Spain), and in 2005 a similar, albeit much smaller, scheme started in the neighbouring Pyrenean country of Andorra. Here we provide a first thorough assessment of butterfly trends in both areas for the last 15 years. Several patterns emerged, above all a worrying decline of a substantial part of the fauna. It was also evident that habitat specialists are experiencing greater declines than habitat generalists, thereby butterfly communities becoming progressively dominated by common species. However, habitat indicators based on characteristic species also revealed that trends are actually associated with habitat types, grassland and scrub specialists declining strongly but woodland specialists showing a marginal increase. These differences are certainly related to profound landscape changes, mainly a dramatic reduction of semi-natural grasslands and open Mediterranean scrub, and a major increase in woodlands. The general effect of climatic warming on butterfly populations was investigated by using the temperature community index (CTI) approach. The thermal structure of butterfly communities remained very stable over time, except in one case where, contrary to the expectations, a significant negative trend in the CTI was noted. However, this surprising result can be explained by taking into account the above-reported pattern of butterfly communities becoming dominated by common species, characterized by low thermal indices in comparison with declining Mediterranean specialists.

Keywords

Butterfly monitoringPopulation trendsLand-use changesHabitat indicatorsClimatic warmingMediterranean basin

Introduction

Following the seminal studies by Thomas (1984, 1991) in the UK, in recent decades interest in the ecology and conservation of butterflies has grown throughout Europe. Worrying trends, with range contractions and population declines, have been reported for many species and habitats in a number of countries, in most cases as a consequence of habitat destruction and fragmentation (van Swaay et al. 2006).

A commonly reported pattern is that of greater declines in habitat specialists than in habitat generalists (e.g. Warren et al. 2001; van Swaay et al. 2006). In the UK, for example, butterfly generalists have remained mostly stable or even experienced a slightly positive trend over the last 30 years (Botham et al. 2008). This phenomenon implies changes in the composition of butterfly communities, which gradually become dominated by common species with fewer habitat requirements to the detriment of species associated with more valuable habitats (González-Megías et al. 2008). Although this seems to be a well established pattern, geographical variability in population trends is also expected to occur because of differences in climate and land use between European countries. For instance, it has been predicted—and already seen to occur—that climate warming will positively affect butterfly species at the northern edges of their European ranges, but negatively affect those at the southern edges of their ranges in the Mediterranean basin (e.g. Warren et al. 2001; Stefanescu et al. 2004, 2010; Menéndez et al. 2006; Wilson et al. 2007; Merrill et al. 2008; Pöyry et al. 2009). However, at the northern margins the extent to which range expansions occur also depends on other factors such as landscape structure (Hill et al. 2001; Menéndez et al. 2006). Moreover, in some countries in central Europe human impact has been so high (e.g. in the form of levels of nitrogen pollution and changes in land use) that even the commonest and most widespread butterflies are undergoing serious declines (Van Dyck et al. 2009).

Unfortunately, the lack of comprehensive datasets on butterfly distribution and population abundance has generally prevented any attempt to go beyond speculation in the case of southern Europe (but see Wilson et al. 2005, 2007, for an exception). From the point of view of butterfly conservation this represents a serious flaw, as the Mediterranean basin is one of the continent’s richest areas for butterflies and with the highest number of endemic species (Dennis and Schmitt 2009). In this paper, we focus on one of the few study systems in Southern Europe for which good butterfly population data exists. In 1994, a Butterfly Monitoring Scheme aimed at systematically recording butterfly populations over an extensive area was launched in Catalonia (NE Spain). This monitoring programme has remained active, and has even increased its geographical coverage and the number of studied butterfly sites; consequently, we are now in a position to perform a first thorough assessment of population trends over the last 15 years for a substantial part of the region’s butterfly fauna. In addition to population trends, we examine the impacts of land-use change and climate change.

Materials and methods

Butterfly data

Data on butterfly abundance were collected between 1994 and 2008 at 95 sites as part of the Catalan Butterfly Monitoring Scheme or CBMS (recording period: 1994–2008, 89 sites) and the Andorran Butterfly Monitoring Scheme or BMSAnd (recording period: 2006–2008, 6 sites) (Fig. 1). Both programmes are based on standardized transect counts of adult butterflies and all use exactly the same methodology (for further details see: www.catalanbms.org/; www.bmsand.ad/) and thus for the purposes of the present analysis data from the CBMS and the BMSAnd were pooled into a single dataset. To ensure a more homogeneous geographical treatment, data from four sites in the Balearic Islands were excluded from the analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9325-z/MediaObjects/10841_2010_9325_Fig1_HTML.gif
Fig. 1

Map of the study region showing the location of the 95 butterfly transects that were used in the calculation of population trends. The size of the circles is proportional to the number of recording years of each sampling site. Also shown are the generally accepted boundaries of the biogeographical regions present in Catalonia

The number of monitored transects increased steadily from 10 in 1994 to 62 in 2008 (Fig. S1). Even though there has been some turnover in the number of sites, an important fraction remained stable throughout the whole recording period (e.g. by 2008, data series of seven or more years were available for 37 transects; Fig. S1). Monitoring sites are situated at an altitudinal range of 0–2,275 m a.s.l. and represent a rich variety of biotopes, including arable farmland, different types of grasslands, shrubland and a variety of forests in coastal and upland areas (Table 1). At each site, counts are conducted along fixed routes (mean length ± SD: 1,675 ± 728 m; range: 728–4,297 m), typically stratified into shorter sections (9 ± 2.5 sections per transect; range: 5–16) to subsample major variations in habitat.
Table 1

Main habitats and plant communities represented in the CBMS and BMSAnd in 1994–2008, with the number of sites at which they appear in

Habitat type and vegetation zone

Dominant plant community

Nr of sites

Mediterranean lowlands

 Holm oak woodland

Coastal holm oak woodland

43

 

Upland holm oak woodland

4

 

Continental holm oak woodland

6

 Scrublands

Rosemary and Erica scrub

2

 

Rosemary and white flax scrub

4

 

Continental holly oak and Rhamnus garrigue

3

 

Coastal holly oak and dwarf fan-palm garrigue

2

 

Thyme and Sideritis scrub

1

 

Continental gypsum and thyme scrub

1

Coast

 Riparian and freshwater vegetation

Coastal wetland communities

7

 Coast

Halophilous communities

3

Sub-Mediterranean and central European-type humid uplands

 Dry deciduous oak and pine woodland

Downy oak and black pine woodland

1

 

Downy oak woodland with box

4

 

Black pine woodland

2

 Humid deciduous oak and beech woodland

Humid deciduous oak and ash woodland

1

 

Pyrenean oak woodland

1

 

Beech woodland

2

 

Ling and dog-violet scrub

1

 Hazel and Scots pine woodland

Scots pine woodland

1

High mountain and subalpine habitats

 Subalpine

Common juniper scrub

1

 

Subalpine grassland

5

Whenever conditions were suitable for butterfly activity, counts were made on a weekly basis starting on March 1 and ending on September 26, a total of 30 recording weeks per year. All butterflies within 2.5 m on each side and 5 m in front of the recorder were counted, as per the standard procedure described by Pollard and Yates (1993). Whenever possible, each transect was always walked by the same person to minimize sampling errors.

Throughout the study period, a total of 170 butterfly species were recorded (see the complete list in Table S1). Of these, a subset of 78 species was used in the analysis of population trends (see below).

Indices of abundance and population trends

At the end of the recording season, indices of abundance were calculated for each of the commonest butterfly species, which were arbitrarily defined as those appearing in an average of seven or more transects per year (see Table S1). As an exception, we excluded a few species whose identification was problematic and often unreliable (e.g. Grizzled Skipper Pyrgus malvoides, Red Underwing Skipper Spialia sertorius, Chapman’s Blue Polyommatus thersites, White Admiral Limenitis camilla, etc.; Table S1), but included a few uncommon species with locally strong populations (e.g. Provence Hairstreak Tomares ballus and Purple-shot Copper Lycaena alciphron) or considered to be of special conservation value (e.g. Large Blue Maculinea arion and Chestnut Heath Coenonympha glycerion; Table S1).

Because of complex phenology involving overlapping generations in most multivoltine species (e.g. Stefanescu et al. 2003), we calculated a single annual index of abundance, rather than separate indices for different generations. Each index included estimated values for missing counts (i.e. values estimated as the mean of the preceding and succeeding counts, as described by Pollard and Yates 1993). The number of missing counts was low throughout the recording period (3.99 ± 2.76 missing weeks per season out of 30 possible); transects with more than 15% of counts missing in any given year were excluded from the analyses.

For each species, indices for different sites were combined into a single national index of abundance using TRIM, a widely used freeware programme for the analysis of series of counts with missing data (e.g. monitoring data), based on loglinear regressions (Pannekoek and van Strien 2005). As is habitual in monitoring programmes based on volunteer work, recording sites were not randomly distributed and so certain areas were oversampled and others undersampled (see van Swaay et al. 2002 for a discussion of this problem). We thus used the weighting procedure implemented in TRIM to correct a sampling bias at county level (average surface area for each county or comarca =767.35 ± 392.92 km2, n = 41), given that at this level environmental conditions remain highly homogeneous. Each recording site was given a weight factor calculated as the quotient between the county’s surface area and the number of recording sites in that county. All weights were expressed in relation to the lowest value, which was re-scaled to 1 (final weights ranged from 1 to 14.8). For TRIM requirements, the counties were nested into a covariable distinguishing between two strata (1-Mediterranean biogeographical region, 2-European, subalpine and alpine biogeographical regions).

Habitat specialisation and population trends

In order to test whether habitat specialists were undergoing more serious declines than habitat generalists, population trends were analysed in relation to the degree of habitat specialisation of each butterfly species. Following Julliard et al. (2006), an index of habitat specialisation (IHS) was calculated based on how the density of each species was partitioned across the habitats most widely represented in the CBMS. Density data were originally obtained from 314 transect sections representing 17 CORINE habitat types (Moss et al. 1990). To avoid pseudoreplication, this number was reduced to 173 independent units by pooling those sections representing the same habitat type in the same transect route. For each species, the IHS was then calculated as the coefficient of variation of the average densities within the 17 CORINE habitat types. The IHS is closely related to the number of occupied habitat classes, with low values for species homogeneously distributed across all habitats and the opposite for species restricted to certain habitat types. To test whether or not population trends are dependent on the degree of habitat specialisation, the slopes of trends as calculated by TRIM were regressed against the IHS values: a significant negative relationship is predicted by the assumption of more serious declines in habitat specialists.

Habitat preferences of butterflies and habitat indicators

In a subsequent step, we investigated whether or not trends in biodiversity were linked to certain habitat types. Following the same methodology as developed for European birds (Gregory et al. 2005), habitat indicators were constructed based on the population trends of characteristic butterfly species. Annual indices rather than the abundances of characteristic species were averaged (using geometric means) to give each species an equal weight in the resulting indicator.

To identify habitat preferences we used species’ density data across the 17 CORINE habitat types as detailed above. In this second analysis, habitat types were grouped into five broad habitat categories (scrub, grasslands, woodlands, wetlands and agricultural/ruderal habitats) and then the average abundance of each species was calculated for each category (expressed as the number of individuals/100 m per year). A species concentrating over 50% of its density in one particular habitat was considered as a specialist or as a characteristic species of that habitat; otherwise, it was considered as a generalist (Table 2). Knapweed Fritillary Melitaea phoebe, Dusky Heath Coenonympha dorus, Silver-washed Fritillary Argynnis paphia and Camberwell Beauty Nymphalis antiopa did not surpass the 50% density threshold in a specific habitat, but were considered as specialists (the first two of grasslands, the latter two of woodlands) following van Swaay et al. (2006) and our knowledge of their ecology in Catalonia.
Table 2

Characteristic species of the broad habitat types used for the habitat indicators

 

IHS

Nr of sites

Scrub specialists

 Lycaena alciphron

2.073

13

 Glaucopsyche melanops

2.067

43

 Thymelicus sylvestris

2.013

34

 Hipparchia semele

1.940

43

 Melitaea cinxia

1.672

35

 Issoria lathonia

1.638

55

 Coenonympha arcania

1.497

51

 Callophrys rubi

1.309

77

 Pyronia tithonus

1.288

58

Grassland specialists

 Aglais urticae

3.127

32

 Erynnis tages

2.299

34

 Hesperia comma

2.180

23

 Polyommatus coridon

2.090

25

 Tomares ballus

2.003

28

 Melanargia occitanica

1.992

21

 Colias alfacariensis

1.779

54

 Coenonympha dorus

1.716

34

 Glaucopsyche alexis

1.705

40

 Polyommatus bellargus

1.553

43

 Melitaea didyma

1.352

58

 Brintesia circe

1.173

69

 Coenonympha pamphilus

1.064

50

 Melitaea phoebe

1.052

64

Woodland specialists

 Neozephyrus quercus

1.543

43

 Nymphalis antiopa

1.444

31

 Pararge aegeria

1.423

90

 Argynnis paphia

1.342

63

 Polygonia c-album

1.299

58

 Celastrina argiolus

1.245

83

 Limenitis reducta

0.959

69

Generalist species

 Lasiommata megera

0.517

92

 Gonepteryx cleopatra

0.635

86

 Colias crocea

0.659

95

 Thymelicus acteon

0.681

66

 Pieris brassicae

0.748

95

 Polyommatus icarus

0.750

92

 Leptidea sinapis

0.802

79

 Melanargia lachesis

0.806

79

 Pieris rapae

0.852

95

 Maniola jurtina

0.861

84

 Lycaena phlaeas

0.893

89

 Gonepteryx rhamni

0.921

77

 Iphiclides podalirius

0.957

87

Also shown is their corresponding Index of Habitat Specialization (IHS, see text) and the number of sites where each species was recorded. Generalist species are those homogeneously distributed between habitat types

No indicators were built for wetlands and agricultural/ruderal habitats, since the number of characteristic species was very low (none and four, respectively). On the other hand, an indicator of habitat generalists was developed to further test the hypothesis that population trends are dependent on the degree of habitat specialisation (see above). Only data for the most generalist species—here defined as those having IHS <1—were used (Table 2). The two long-distance migrants Red Admiral Vanessa atalanta (IHS = 0.489) and Painted Lady Cynthia cardui (IHS = 0.668) were excluded, as their population dynamics are totally dependent on the environmental situation in their areas of origin (Pollard et al. 1998; Stefanescu 2001).

Temporal trends of habitat indicators were tested by linear regression, with years as the independent variable.

Impact of climatic change on butterfly populations

To test for a general effect of climatic change on butterfly populations, the community approach developed by Devictor et al. (2008, in prep.) was used. Selected butterfly communities were characterized by their community temperature index (CTI) for each year from 1994 to 2008. The CTI has been defined as the average of the long-term mean temperature within the range of each species occurring in the assemblage, weighted by species abundance (Devictor et al. 2008). A direct response of butterfly species to climatic warming would lead to a predictable increase in the CTI, given that populations adjust their densities according to each species-specific temperature requirement. We tested for temporal trends in CTI for a subset of six monitored sites that had complete data from 1994 to 2008. These sites encompass an altitudinal range of 0–1,100 m and thus include a variety of environmental types (from coastal sites to lowland and upland sites dominated by different kinds of scrub and forests).

Results

A summary of the population trends recorded between 1994 and 2008 in Catalonia and Andorra is given in Table 3. Almost half of the studied species (37 out of 78 species, or 47.4%) showed significant population trends, with negative trends clearly predominating over positive trends (28 vs. 9 species). Furthermore, 10 trends were classified by TRIM as steep declines, compared to only four strong increases. Among the species with uncertain trends, negative values were also predominant (13 vs. 6 species), suggesting that a general pattern of population declines in the region will be further confirmed as more data become available in future years. Only 22 species (28%) remained stable throughout the whole period.
Table 3

Population trends of 78 butterfly species in Catalonia and Andorra based on 15 years (1994–2008) of transect count data, as calculated with TRIM

Species

Overall trend

Stable trend

 

 Carcharodus alceae

−0.014

 Thymelicus sylvestris

0.0051

 Thymelicus acteon

−0.0089

 Papilio machaon

−0.0176

 Anthocharis cardamines

−0.0005

 Euchloe crameri

0.0022

 Pontia daplidice

0.0052

 Callophrys rubi

−0.0164

 Satyrium esculi

0.01

 Lampides boeticus

−0.0091

 Cupido alcetas

0.021

 Celastrina argiolus

−0.0038

 Inachis io

0.012

 Polygonia c-album

0.0011

 Limenitis reducta

−0.0047

 Charaxes jasius

−0.0081

 Pararge aegeria

−0.0046

 Lasiommata megera

−0.0014

 Coenonympha pamphilus

0.0081

 Pyronia bathseba

−0.009

 Maniola jurtina

0.0087

 Melanargia lachesis

−0.0129

Uncertain trend

 

 Hesperia comma

−0.0316

 Zerynthia rumina

−0.0416

 Anthocharis euphenoides

−0.034

 Aporia crataegi

0.0255

 Neozephyrus quercus

0.0332

 Tomares ballus

−0.0239

 Cacyreus marshalli

−0.0584

 Cupido minimus

−0.0784

 Pseudophilotes panoptes

−0.0446

 Glaucopsyche alexis

−0.0243

 Glaucopsyche melanops

−0.0614

 Aricia cramera

−0.0417

 Polyommatus coridon

−0.014

 Boloria dia

−0.073

 Melitaea phoebe

0.0257

 Melitaea didyma

0.0021

 Melitaea deione

0.0481

 Hipparchia statilinus

−0.0251

 Hipparchia fidia

0.0356

Decline (Moderate)

 

 Erynnis tages

−0.0786

 Ochlodes venatus

−0.0295

 Iphiclides podalirius

−0.024

 Leptidea sinapis

−0.0367

 Pieris brassicae

−0.028

 Pieris napi

−0.0437

 Colias crocea

−0.0241

 Colias alfacariensis

−0.0786

 Lycaena phlaeas

−0.0342

 Lycaena alciphron

−0.0501

 Leptotes pirithous

−0.0334

 Polyommatus semiargus

−0.0787

 Polyommatus icarus

−0.0264

 Polyommatus bellargus

−0.0805

 Issoria lathonia

−0.0304

 Vanessa atalanta

−0.012

 Euphydryas aurinia

−0.0696

 Pyronia tithonus

−0.0202

(Steep)

 

 Maculinea arion

−0.398

 Polyommatus escheri

−0.1498

 Cynthia cardui

−0.0792

 Aglais urticae

−0.1041

 Melitaea cinxia

−0.1061

 Coenonympha arcania

−0.0825

 Coenonympha glycerion

−0.1725

 Coenonympha dorus

−0.103

 Pyronia cecilia

−0.1241

 Melanargia occitanica

−0.2259

Increase (Moderate)

 

 Pieris rapae

0.0389

 Gonepteryx rhamni

0.0459

 Argynnis paphia

0.0408

 Hipparchia semele

0.048

 Brintesia circe

0.0678

(Steep)

 

 Gonepteryx cleopatra

0.106

 Libythea celtis

0.1952

 Nymphalis antiopa

0.1921

 Nymphalis polychloros

0.1355

The overall trend is the slope of the regression line over the whole study period, based on imputed indices with intercept. The 95% confidence interval of a trend estimate is computed by multiplying the standard error by 1.96. If this interval does not include the value 1, then the trend is statistically significant. Also shown is the classification of the trend estimate given by TRIM for a 20-year period. Declines or increases refer to significant trends, while stable or uncertain trends are non-significant

There was a significant relationship between population trends and habitat specialization (R2 = 0.18), with specialists (i.e. species with high IHS values) generally showing steeper declines than generalists (Fig. 2). However, the analysis of habitat indicators clearly indicated that not all habitat specialists were declining and that their fate was actually associated with certain habitat types (Fig. 3). Thus, although a grassland and a scrub indicator summarizing broad trends experienced by characteristic species revealed a strong decline between 1994 and 2008 (grassland indicator: F = 11.41, P = 0.0049; scrub indicator: F = 14,66, P = 0.0021), a woodland indicator showed a marginal increase (F = 4.55, P = 0.053). By 2008, the indices for grassland and scrub specialists were 67 and 50% lower than the base year in 1994, respectively, while the indicator for woodland specialists was 115% higher (Fig. 3). Interestingly, the overall abundance of generalist species (i.e. those with no clear habitat preferences) remained very stable (F = 0.00, P = 0.98), thereby confirming the pattern—that common and widespread butterflies are to some extent buffered against population declines—shown in Fig. 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9325-z/MediaObjects/10841_2010_9325_Fig2_HTML.gif
Fig. 2

The relationship between population trends and an index of habitat specialization (IHS) in the 78 studied butterfly species. The highly significant negative correlation between both variables (r = −0.421, P < 0.0001) reveals a trend for greater population declines in habitat specialists than in habitat generalists

https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9325-z/MediaObjects/10841_2010_9325_Fig3_HTML.gif
Fig. 3

Temporal trends in habitat indicators based on the characteristic species of broad habitat types (woodland: 7 species; scrub: 9 species; grassland: 14 species; generalist: 13 species). Grassland and scrub indicators declined significantly, while the woodland indicator marginally increased between 1994 and 2008. By contrast, the generalist indicator remained stable over the same period

On the other hand, no clear pattern emerged from the analysis of the CTI (Fig. 4). Only in one out of six butterfly communities analysed was there a significant negative trend in the CTI. In this case, contrary to expectations under a scenario of climatic warming, the community became progressively dominated by cold species. Apart from this exception, the thermal structure of butterfly communities remained very stable over time. The lack of significant trends cannot be attributed to a bias in the choice of the sampling sites since they included a wide range of altitudes and ecological conditions (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9325-z/MediaObjects/10841_2010_9325_Fig4_HTML.gif
Fig. 4

Trends in the Community Temperature Index (CTI) of six butterfly communities between 1994 and 2008. Only Can Ferriol showed a significant negative trend (b = −0.082, F = 14.94, P = 0.002). The altitudes and main habitat types of each sampling site are as follows: El Cortalet, 0 m a.s.l.; Can Ferriol: 200 m; Darnius: 200 m; Can Liro: 330 m; Can Jordà: 600 m; El Puig: 1,100 m

Discussion

Our study is the first to provide a general quantitative assessment of recent butterfly population trends in a large region in the Mediterranean basin (NE Iberian Peninsula). On the basis of 15 years of transect data, we have shown that in nearly half of the species significant trends are occurring, consisting mainly of population declines. Although the assessment was based on a subset of the species known in the region (78 out of 200 species), our conclusions can be probably applied to the whole of the region’s butterfly fauna. Indeed, the analysis was mostly focused on common butterflies widely distributed across the country, that is, species characterised by a relative ecological flexibility and thus more resistant to environmental change. Unfortunately, data for rare species are generally too scarce to be used in any comparable analysis. However, even with the available dataset, we obtained clear evidence that the species showing the most serious declines are true habitat specialists (e.g. the uncommon and locally distributed species) (Fig. 2). As an example, the steepest declines were recorded for Maculinea arion, Western Marbled White Melanargia occitanica and Coenonympha glycerion, all three scarce butterflies with narrow ecological requirements and very local populations in Catalonia and Andorra. Thus, if the whole pool of species, including even the rarest ones, could be analysed, it is highly likely that butterfly declines would appear to be even more widespread.

Moreover, our finding further confirms the general phenomenon whereby butterfly communities are becoming more dominated by generalist species (Polus et al. 2007; González-Megías et al. 2008), which has been described in countries such as Belgium and the UK subject to more severe processes of landscape intensification. This is a particularly worrying result in a region widely recognized as a world biodiversity hotspot (Blondel and Aronson 1999; Myers et al. 2000) in which a high concentration of endemic butterfly species occurs (Dennis and Schmitt 2009).

The analysis of habitat indicators provides some clues as to which factors are responsible for the above-reported pattern. Our data show very clearly that population trends differ widely depending on the habitat type under consideration: while the most characteristic species of open habitats such as different kinds of grasslands and scrub have undergone very rapid declines in recent years, an opposite, albeit weaker trend, has been observed for woodland species. This is certainly related to the profound changes in the landscape that have occurred in the country—and, more generally, in vast areas of the northern Mediterranean basin—in recent decades. Amongst the most evident are the abandonment of traditional land uses such as grazing and mowing in low productive upland areas (e.g. Preiss et al. 1997; Debussche et al. 1999) and pressure from increasing agricultural intensification and urbanization in fertile lowlands and coastal areas (e.g. Petit et al. 2001; Pino et al. 2009). Conjointly, these opposing forces are leading to a dramatic reduction of semi-natural grasslands and open Mediterranean scrub, and also to a major increase in woodlands, especially in mountain areas. In a mere 100 years woodland cover in Catalonia has more than doubled, increasing from 600,000 to 1,400,000 ha, while in the short term between 1987 and 1997 the increase has been estimated to be in the order of 15,000 ha (Brotons et al. 2004).

These large scale landscape changes have probably seriously affected the fauna and flora. We hypothesize that woodland butterflies benefit from the increase in forest cover, just as has been observed in Catalan birds (Brotons et al. 2004). The situation in open habitats is radically different and bears a close resemblance to that in central and northern Europe (e.g. Wallis de Vries et al. 2002; Krauss et al. 2010). For butterflies in particular, over the last 15 years there has been a strikingly similar decrease in the average abundance of grassland species in our region and in western Europe in general (van Swaay et al. 2008).

Aside from habitat preferences, other ecological correlates could also be influencing the direction of population trends, as has been shown by other Lepidopteran studies (e.g. Koh et al. 2004; Mattila et al. 2006; Pöyry et al. 2009). A formal analysis considering different life-history traits is beyond the scope of this paper, but could be the subject of a future study. However, it is interesting to note that population trends differed significantly between overwintering strategies (F3,67 = 5.90, P = 0.0012), with species overwintering as adults (except the Small Tortoishell Aglais urticae) showing comparatively stronger increases than the others. Undoubtedly, this remarkable finding deserves further investigation.

One may also expect climatic warming to play an important role in population trends, above all in view of the compelling evidence highlighting this factor as one of the main drivers of change in butterfly communities currently at work (e.g. Menéndez et al. 2006; Wilson et al. 2007; Pöyry et al. 2009). This possibility seems actually more likely given the theoretical work that predicts that climatic warming will lead to losses in butterfly species richness in the study region (Stefanescu et al. 2004, 2010). Recently, model projections of change in potential range extent have been used as a proxy for the impact of climate change on butterfly species at a continental scale (Heikkinen et al. 2010). Unfortunately, climatic envelope models based on high resolution data are not available for butterfly species at the scale of the study area and so such an approach in our case would be inapplicable. Instead, here we have used the temperature community approach developed by Devictor et al. (2008, submitted) to test for a general effect of climatic change on butterfly populations.

Quite surprisingly, none of the six communities for which data were available from 1994 to 2008 showed an increase in temperature. In fact, the only significant trend was in the opposite direction, that is, one community apparently became dominated by colder species during the study period. Although our results contradict previous findings for bird and butterfly communities in various central and northern European countries (Devictor et al. 2008, submitted), we believe that this apparent abnormal pattern can be explained by taking into account some confounding factors. In particular, land use changes may compensate for the expected increase in CTI by favouring butterfly generalists and, conversely, by being detrimental to butterfly specialists. The former species are typically more widespread in Europe, which means that their Species Temperature Index or STI (i.e. an index summarising the long-term average temperature over the species’ entire European range) are usually much lower than the STI of Mediterranean specialists inhabiting dry meadows and sclerophyllous scrub in southern latitudes. The CTI of a butterfly community for a given year at a given site is the average of each STI weighted by the species’ abundance. Therefore, the above-reported phenomenon of butterfly communities becoming progressively dominated by generalists would lower the CTI values in most cases and thus act in the opposite direction to climatic warming.

Interestingly, the only site where a significant trend in the CTI was recorded is situated in the metropolitan area of Barcelona in an area subject to high levels of human disturbance (e.g. human frequentation and habitat fragmentation). This continuous pressure results in the deterioration of natural habitats such as dry meadows and evergreen oak forest, and the expansion of ruderal habitats (e.g. Guirado et al. 2007). As a consequence, Mediterranean endemics such as Southern Gatekeeper Pyronia cecilia, Melanargia occitanica and Dusky Heath Coenonympha dorus have experienced dramatic declines in the last two decades, while generalist species such as Small White Pieris rapae and Speckled Wood Pararge aegeria have become considerably more common. These changes have led to a substantial decrease in the CTI of this butterfly community.

Although not revealed by our analysis, other data strongly suggest that climatic warming is indeed influencing Mediterranean butterfly communities. For instance, several highly mobile African species have expanded their ranges recently and have started to be increasingly recorded in our region (e.g. Plain Tiger Danaus chrysippus, Monarch Danaus plexippus, Desert Orange Tip Colotis evagore and Green-striped White Euchloe belemia), while remarkable advances in the flight periods of different species have also been recorded at a number of sites (Peñuelas et al. 2002; Stefanescu et al. 2003; Gordo and Sanz 2005). Moreover, detailed community and auto-ecological studies carried out in a mountain range in central Spain have convincingly shown altitudinal shifts in the predicted direction and also provide a number of mechanistic processes that help explain the observed patterns (Wilson et al. 2005, 2007; Merrill et al. 2008).

We can therefore conclude that even though the approach of Devictor et al. (2008, submitted) has been successfully used for detecting rapid responses of birds and butterflies to climatic warming, its usefulness may be constrained by other factors such as land use changes that have a potentially greater impact on the structure of natural communities.

Acknowledgments

Thanks are due to all the recorders who contribute data to the Catalan Butterfly Monitoring Scheme and BMSAnd. The Butterfly Monitoring Scheme in Catalonia is funded by the Departament de Medi Ambient i Habitatge de la Generalitat de Catalunya. The Diputació de Barcelona, Patronat Metropolità Parc de Collserola and Fundació Caixa de Catalunya have also given financial support to this project. The BMSAnd project is supported by the Centre d’Estudis de la Neu i la Muntanya d’Andorra (CENMA), from the Institut d’Estudis Andorrans (IEA).

Supplementary material

10841_2010_9325_MOESM1_ESM.doc (467 kb)
Table S1Species detected in the CBMS, with the minimum, maximum and average number of sites at which they were annually recorded. Taxonomy as per Karsholt and Razowski (1996), with slight variations. Karsholt O, Razowski J (1996) The Lepidoptera of Europe. A Distribution Checklist. Apollo Books, Stenstrup. (DOC 467 kb)
10841_2010_9325_MOESM2_ESM.doc (26 kb)
Fig. S1(a) Number of sites where butterflies were recorded each year; (b) Distribution of the complete annual series available (1994–2008) for all the sites used for calculating butterfly trends. (DOC 25 kb)

Copyright information

© Springer Science+Business Media B.V. 2010