Butterfly diversity and biogeography on the Croatian karst mountain Biokovo: Vertical distribution and preference for altitude and aspect?

Geographic isolation, altitude, climate, landscape and habitat are significant predictors of butterfly diversity in mountain ecosystems. Their diversity and its dependence on altitude, aspect (compass bearing) and biogeographic characteristics of the butterflies were surveyed on the karst mountain Biokovo in southern Croatia. The results affirm that there is a high diversity of butterflies in the study area and the species composition and biogeographic elements are more dependent on altitude than aspect of the mountain. The present study indicates that climate, relief and habitat preferences strongly influence the biogeographic features of species and the relationship between species richness per site and altitude, aspect and the altitude-aspect interaction.

In mountain ecosystems, the distribution of species is determined by the suitability of habitat and climate (Storch et al., 2003).The main factors that influence species distribution and diversity include geographic isolation, altitude, climate and features of landscape and habitat such as structure, heterogeneity and quality (e.g.Kudrna, 1986;Fleishman et al., 1998Fleishman et al., , 2003;;Wettstein & Schmid, 1999;Pyrcz & Wojtusiak, 2002;Storch et al., 2003).
Mountain ecosystems, though considered "simple" because the number of species decreases with altitude, are extremely heterogeneous (Haslett, 1997a).The diversity and area/size of habitats at different altitudes affects spatial distribution, species richness, morphology, physiology, life cycle and behavioural patterns (Haslett, 1997b).
Research on the diversity of butterflies occurring on the karst mountain range Biokovo (Dalmatia, southern Croa-tia) has continued, off and on, since the 19 th century (Mann, 1869;Stauder, 1923;Sijari , 1991a, b;Mihoci & Šaši , 2005).This mountain is a part of the Dinaric karst, which is a mountain range approximately 700 km long and 80-210 km wide that stretches from the Julian Alps along the border between Italy and Slovenia, passes through the Croatian coast and islands, western and southern Bosnia and Herzegovina, and ends in south eastern Montenegro and north western Albania (Tišljar et al., 2002).Mt.Biokovo is closest to the Adriatic Sea (Fig. 1) and as a consequence there are areas on the mountain with continental, montane or Mediterranean climates and vegetation.Certain habitat characteristics of Mt.Biokovo cannot be found elsewhere in the Dinaric karst.To protect these features, the habitat diversity and species listed in the Habitats Directive 92/43/EEC annexes (European Comission, 1992; State Institute for Nature Protection, 2008), Biokovo was declared a Nature Park in 1981 and proposed as a NATURA 2000 site in 2010.
This study determines the influence of altitude and topographic aspect on the distribution and biogeographic characteristics of butterflies on Mt.Biokovo.It also compares the composition and biogeographic characteristics of the species on Mt.Biokovo with those on several other Dinaric karst mountains (Sket, 1994): Mt.Velebit (Croatia) (Mihoci et al., 2007), Mt.Durmitor (Montenegro) (Sijari et al., 1984), Mt.U ka (Croatia) (Rebel, 1910;Ku ini , 2010) and the complex of Mt. vrsnica and Vran (Bosnia and Herzegovina) (Sijari , 1996;Ku ini et al., 2005).Since similar faunistic research has been done on neighbouring karst areas in the Balkans it was possible to carry out a cluster analysis of the results of these and our study on Biokovo.

Study sites
Mt. Biokovo is 36 km long and 9.5 km wide (Fig. 1).It has an oblong shape and is orientated along a northwest-southeast axis.The highest peaks are in the southeast of the range, with the tallest Saint Jure peaking at an altitude of 1762 m (Kušan, 1969).Mt.Biokovo is isolated from other mountain areas in the Dinarids, as it adjoins the Adriatic Sea to the south and is separated from the nearby mountains in Croatia and Bosnia and Herzegovina by a broad lowland plain in the north.
Geomorphologic features of Mt.Biokovo originated during the karstification process in fluvial periods.Bedrock originated from sedimentation, and the main geomorphologic characteristics are due to limestone and dolomite karst, and the occasional water regime (Kušan, 1969;Telbisz et al., 2009).The aspect of the sea-facing side of the mountain is much steeper than that of the mainland side.In mid-range altitudes between 400 and 1100 meters, the sea-facing aspect is composed predominantly of cliffs and steep rocks, features which are absent on the mainland side of the mountain.
The vegetation on Mt.Biokovo includes both older Mediterranean and newer Boreal and European forms, together with different types of xerophytic, mesophytic and submontane vegetation, in which Mediterranean elements dominate.Its geographical isolation has made Mt.Biokovo a hotspot of plant endemism, with over 30 endemics (Šoli , 1983).

Butterfly recording
Butterflies were recorded by standard methods using an entomological net in all the vegetation periods from 1994 to 2004 at 10 sites on the mainland side (northeast aspect) and 9 on the coastal side (southwest aspect) and at the top of St. Jure (N/S) (Fig. 1).The sites sampled at the different altitudes were determined exclusively by the presence of meadow/grassland, especially on the forested northern aspect.In several analyses, localities from the northern (N) and southern (S) aspects were grouped into three altitude zones (Z) (zone 1, NZ1 & SZ1: 0-500 m; zone 2, NZ2 & SZ2: 500-1000 m; zone 3, NZ3 & SZ3: above 1000 m).The same amount of time was spent sampling all the sites visited on five occasions between April and September, but in different years: the southern aspect was visited between 1994 and 2000 and the northern aspect between 2000 and 2004.Sampling at each site was done over a 1-h period, and all butterflies seen and captured in that time were recorded.Butterflies were recorded only on dry, sunny and warm days between 10 a.m. and 5 p.m. (Pollard & Yates, 1993).

Data processing
All the tests below were computed separately for altitude and aspect.The number of species at each locality and the mean altitude for each species were determined on both the northern and southern aspects.Change of species richness with altitude and aspect was demonstrated by a regression-based test.Alpha diversity was calculated using the Shannon-Wiener diversity index (H) and the Simpson diversity index (D).The Shannon-Wiener index is more sensitive to rare species, but its numerical value is less meaningful than some other indices.In contrast, the Simpson index is relatively insensitive to rare species, making it relatively stable with respect to sample size.It can be used to count infinite and finite populations, as well as binary variables (Whittaker, 1972).
Species turnover rate was determined according to Schoener (1986) and statistical significance demonstrated by a regressionbased test.
Faunistic similarity between mountain aspects and altitude zones were calculated using the Sørensen similarity index (S) (Sørensen, 1948).Indicator species analysis (ISA) (Dufrêne & Legendre, 1997) was used to determine which species are characteristic of aspects or zones (in two groups).Statistical significance of the indicator values (IVs) obtained was tested using a permutation test with 1000 random permutations.
The effect of altitude or aspect on species composition was tested using variance decomposition with score files obtained through simple, between-and within-groups normalized principal component analysis (PCA) (Dolédec & Chessel, 1991).In addition, three generalized linear models (GLM) (McCullagh & Nelder, 1989;Lepš & Šmilauer, 2003) relating species number per site (SNS) to altitude; aspect and the altitude-aspect interaction were computed using a Poisson log model.
Butterfly diversity was compared using the Bray-Curtis distance for samples collected at five sites: Mt.Biokovo, Mt.U ka, Mt.Velebit, Mt.Durmitor and the complex of Mt. vrsnica and Vran.
Statistical analyses and ISA were conducted using PCOrd-4.37 and PRIMER6 software, whereas inertia decomposition and PCA were conducted using ADE-4 software (Chessel et al., 1998).

RESULTS
A total of 102 species were recorded (Appendix).The most interesting records were the endemic Zerynthia cerisy dalmacijae and Proterebia afra dalmata, and rare Papilio alexanor, all strictly protected taxa in Croatia (Ministry of Culture of Republic of Croatia, 2009).Of the species of Erebia, which all typically occur on mountains, only Erebia melas was recorded.
Higher species richness and higher average number of species per locality (species density) were recorded on the northern than on the southern aspect (88 vs. 80, 18.4 vs. 17.2), but these differences are not statistically significant (t-test, p > 0.05).Similarly, the Shannon (H) and Simpson (D) diversity indices were higher on the northern aspect (Table 1).Sørensen's similarity between the two aspects was relatively high (S = 78.57%),with 66 species shared by the two aspects.Of all the species observed, 22 were recorded only on the northern and 14 only on the southern aspect.
The highest number of species occurred in the lowest altitude zone, zone 1 (total 79, 19.1 per locality).Altitude zones 2 and 3 had a nearly equal number of species (Table 2).The Sørensen similarity index between NZ1 and SZ1 was 56.36%, with 31 species shared by the two zones; 35.00% between NZ2 and SZ2; and 52.75% between NZ3 and SZ3, with 24 species shared.Altitude zones 1 and 2 on the northern aspect were the most similar in species composition, with a Sørensen similarity index of 71.43% (Table 3).
At the highest mean altitude on the southern aspect, 38 species were recorded, compared to 22 at the highest mean altitude on the northern aspect (Appendix).
On both aspects, species turnover rate (STR) increased with altitude (Fig. 3).The calculated rates indicate that a 95% change in species composition is reached at 1052-1550 m on the northern aspect.On the southern aspect, the highest percentage change (81%) is reached at 1600-1762 m.In general, STR is lower on the southern than on the northern side of the mountain (Fig. 3 3. Sørensen similarity index (%) between altitude zones on the northern and southern aspects of Mt.Biokovo (NZ1 -0-500 m on the northern aspect, NZ2 -500-1000 m on the northern aspect, NZ3 -above 1000 m on the northern aspect, SZ1 -0-500 m on the southern aspect, SZ2 -500-1000 m on the southern aspect, SZ3 -above 1000 m on the southern aspect).regression-based test showed statistical significance for STR on the northern (y = 0.022x + 0.72) (p < 0.001) and insignificance on the southern side (y = -0.0002x+ 0.759) (p > 0.05).Aspects when combined are highly statistically different (t-test, p < 0.001) (Fig. 4a, b).Variance decomposition analysis showed that the effect of altitude on the variability in the species composition of butterflies is 2-fold greater than the effect of aspect (Table 4).
The General Linear Modelling (GLM) analysis using the model comparing species number per site SNS (predicted vs. observed) with altitude showed that sites S3, N4, N5 and N9 have lower species richness than predicted (below confidence limits), while sites S6, S8, N1 and N8 have higher richness than predicted (Fig. 5a).In the model comparing SNS with aspect, sites N1, N8 and S2 have higher species richness than predicted (Fig. 5b).In the model comparing SNS with the altitude-aspect interaction, sites S3, S9, N4, N5 and N9 have lower species richness than predicted, while sites N1, N8, S2, S6, and S8 have higher richness than predicted (Fig. 5c).
Comparison of biogeographic proportions of species from Mt. Biokovo, Mt.Velebit (Croatia) and Mt.Durmitor (Montenegro) revealed that Circum-Palearctic species accounted for the highest share of species (< 35%), while the proportion of North and Pontic-Mediterranean and Pontic-Mediterranean Oreal species was higher on Mt.Biokovo than on Mt.Velebit or Mt.Durmitor.The share of Sibiric species was several times higher on Mt.Velebit and Mt.Durmitor than on Mt.Biokovo.No Euro-Alpine species were recorded on Mt.Biokovo, in contrast to the other two mountains (Table 6).
vrsnica and Vran is more similar to Mt. Velebit and Mt.Durmitor than to that on Mt.Biokovo or Mt.U ka (Fig. 6).

DISCUSSION AND CONCLUSIONS
The position of Mt.Biokovo on the Adriatic Sea coast, at the border between the continental and Mediterranean climate, as well as its isolation from surrounding mountains, has resulted in a high degree of endemism (Šoli , 1983) and great diversity of habitats (Kušan, 1969).The 627  present study indicates that 53% of the Croatian butterfly fauna occurs on Mt.Biokovo (e.g.Lorkovi , 1989;Mihoci et al., 2005Mihoci et al., , 2006;;Perkovi , 2006;Mihoci & Šaši , 2009;Polak, 2009).The butterfly fauna on the southern and northern slopes is similar, probably because species ranges overlap (S = 78.57%)as the northern and southern localities are relatively close together.On the other hand, the Sørensen index indicated differences in the composition of the fauna in different altitude zones, in particular the isolation of species at mid-altitudes on the southern aspect.
On the southern aspect, more species were found at the highest mean altitude, while more species were found at the lowest mean altitude on the northern aspect.These results are expected if the vertical species distribution is a response to climatic conditions (Settele et al., 2008, Van Swaay et al., 2010), which are hotter on the southern than on the northern aspect (Kušan, 1969;Penzar & Penzar, 1995).Therefore, some species may be highly threatened by recent climate change, such as the high-mountain species Erebia melas or Coenonympha rhodopensis.Some mountain (e.g.E. melas, C. rhodopensis) and lowland species (Zerynthia cerisyi dalmacijae, Z. polyxena, Leptotes pirithous, Lampides boeticus) were found only at high and low altitudes (Appendix).
Butterfly diversity was higher on the northern aspect due to the predominance of suitable small, calcareous, semi-natural grasslands and forest edge habitats.The richest plant and insect communities in Europe occur in calcareous grasslands (Wallis de Vries et al., 2002).The southern aspect is dominated by cliffs and rocky slopes with a scant and specific vegetation comprised mostly of chasmophytes (Kušan, 1969;Mihljevi , 1995;Ridanovi & Šimunovi , 1995).These habitats are less suitable for butterflies, since few of the species of plants occurring there are larval host-plants or a source of nectar for the adults.In contrast to the present study, a peak in butterfly diversity is recorded at mid-altitudes in mountain areas that do not have the specific geomorphology and climate of Mt.Biokovo (Gutiérrez, 1997;Pyrcz & Wojtusiak, 2002).As pointed out by Hodkinson (2005), maximum species richness of terrestrial insects may occur at the highest, lowest or middle altitude, or show no trend with altitude.
The increase in species turnover rate with altitude on the southern aspect probably reflects changes in habitat structure: Aleppo pine forests at the lower altitudes gradually give way to steep chasmophytic cliffs and rocks, while the higher altitudes are rich in mountain grasslands and edge habitats associated with black pine and beech forests (e.g.Kušan, 1969).The greatest change in the fauna on the southern aspect occurred at 1600-1762 m (sites S9-S/N).Site S9 is located between two narrow sinkholes and is an area subject to a high anthropogenic influence (mountain hut, tourists, road vicinity).Site S/N, at the top of the mountain, features wide, open calcareous slopes rich in vegetation the composition of which is influenced by that of the vegetation on both aspects of the mountain and is least affected by human disturbance.Similarly, the highest number of species on the southern aspect occurred at S1 (33) and S2 (30), where meadows are rich in nectar plants and wellpreserved in a botanical reserve.High butterfly richness was also recorded at sites S6 and S8, which have habitat characteristics similar to those of S1 and S2, and confirmed by the GLM analysis using models relating species richness to altitude and the aspect-altitude interaction.
On the northern aspect, a 95% change in species composition occurred at 1052-1550 m (N9-10).In fact, the change in species composition is significant, since there are several high-mountain species at site N10 and only a few at N9.The two sites also differ in terms of altitude and habitat composition: N9 is an isolated meadow surrounding several ponds, whereas N10 is a wide open grassy slope above a beech forest.
On Mt.Biokovo butterfly diversity and composition changes almost twice as much with altitude as with aspect.Numerous authors have pointed out that altitude is one factor that determines climate patterns and changes in floristic composition, both of which strongly affect species distribution (Wettstein & Schmid, 1999;Pyrcz & Wojtusiak, 2002;Storch et al., 2003;Eyre et al., 2005).
The present study revealed that several species, and two in particular, can be used as indicators.On the northern aspect, the endemic P. afra dalmata is an indicator species of mid-altitude regions with extensive semi-natural grasslands and edge habitats, rich in the larval-host plant Festuca spp.and Juniperus bushes for feeding (Mihoci & Šaši , 2005).This taxon lacked suitable habitats at all of the sites on the southern aspect.
Erebia melas is an indicator species of the highest altitude zone.This xeromontane species is the only Erebia recorded on Mt.Biokovo and similarly on Mt.Olympus in Greece (Varga, 1975;Sijari et al., 1984;Mladinov & Lorkovi , 1985;Mihoci et al., 2007).On both mountains E. melas occupies a wide range of rocky habitats, usually up to the highest alpine levels (Varga & Varga-Sipos, 2005).In contrast, Mt.Velebit (Croatia), the longest mountain on the Dinaric karst, has 11 of the 12 Erebia species occurring in this country (Mihoci et al., 2007).E. melas on Mt.Biokovo defines the southern edge of Erebia species distribution in Croatia (Mladinov & Lorkovi , 1985;Jakši , 1988).The absence of other Erebia species on Mt.Biokovo is attributable to the limited presence of high-mountain meadows rich in grasses.These are replaced by rocky slopes with sparse grasses and an unfavourable climate with long and intense summer droughts, high temperatures and infrequent precipitation (Kušan, 1969;Penzar & Penzar, 1995).The "Erebia favourable" habitat occurs on Mt.Velebit where there are 11 species of Erebia and on Mt.Durmitor in Montenegro where there are 12 species.These mountains have a favourable climate and widespread alpine and subalpine meadows and semi-natural grass-covered slopes (Sijari et al., 1984;Forenbacher, 2001;Nikoli , 2004).
The composition of the fauna revealed by biogeographic analysis was expected to reflect the geographical position, climate and ecological characteristics of Mt.Biokovo (Kušan, 1969;Penzar & Penzar, 1995).As expected West-Palearctic, North and Pontic-Mediterranean, Atlantic-Mediterranean and Paleotropic species were predominant on the southern aspect of the mountain, which experiences a Mediterranean climate.On the northern aspect, in contrast, Circum-Paleartic, Pontic-Caspic-Turkestanic-Southsibiric, Sibiric and Pontic--Mediterranean Oreal species are predominant.The finding that Mt.Biokovo has a more similar fauna to Mt. U ka (68%) than to other Dinarid mountains demonstrates the influence of the Adriatic Sea, which results in a Mediterranean climate on the coastal slopes and a continental one on mainland slopes.Both Mt.Biokovo and Mt.U ka lack high-mountain species, such as Coenonympha gardetta Prunner, Lycaena candens Herrich-Schaffer, Polyommatus eros Ochsenheimer, Polyommatus pyrenaicus Boisduval and Erebia spp.(Mann, 1869;Rebel, 1910;Stauder, 1923;Sijari , 1991a, b;Ku ini , 2010) unlike Mt.Velebit, Mt.Durmitor or the complex of Mt.
Comparison of the biogeographic characteristics of Mt.Biokovo, Mt.Velebit and Mt.Durmitor suggests that altitude and the continental and alpine climate accounts for the great similarity in composition of species of butterfly on Mt.Velebit and Mt.Durmitor (78%).The greater influence of continental and alpine climate depends on the geographical position and greater heights of Mt.Velebit and Mt.Durmitor compared to Mt. Biokovo (Sijari et al., 1984;Forenbacher, 2001).
The most common species on all three mountains are Circum-Palearctic species, which is consistent with their dominance in the European fauna (Varga, 1977).However, there is a several-fold lower number of Sibiric species on Mt.Biokovo than on Mt.Velebit and Mt.Durmitor.Even more interestingly, there are no Euro-Alpine species on Mt.Biokovo whereas there are eight on Durmitor.These differences in biogeographical proportions clearly reflect differences in the geographical positions, altitude and habitat characteristics of the mountains.
In conclusion, the present study indicates that the composition and biogeographic features of butterfly species can vary significantly, even over a small area such as the southern and northern aspects of one mountain.This variation reflects the diversity of habitats and the changes in climate and geomorphology with altitude and aspect.

Fig. 2 .
Fig. 2. The species richness at different altitudes on the northern (N -dashed line) and southern (S -solid line) aspects.

Fig. 3 .
Fig. 3. Species turnover rate (STR) between different and increasing altitudes on the northern (N -dashed line) and southern (S -solid line) aspects.13.07 11.76 1.31 Altitude 13.07 11.23 0.74 Aspect Total inertia Within groups inertia Between groups inertia TABLE 4. Inertia decomposition analysis -used to test the prominence of altitude or aspect effect on variability in species composition.

Fig. 5 .
Fig. 5.The GLM analysis.A -model comparing observed with predicted species number at sites at different altitudes on the mountain; B -model comparing observed and predicted number of species at sites on the northern and southern aspects of the mountain; C -model comparing observed and predicted numbers of species at sites with altitude-aspect combined.

TABLE 1 .
Total number of species (TSN), average number of species recorded at the sample sites (ASN) on the northern and southern aspects and the Shannon's (H) and Simpson's (D) diversity indices for the northern and southern aspect of Mt.

TABLE 2 .
Total species number (TSN), average species number recorded at sampling sites (ASN) at different altitude zones and the Shannon's (H) and Simpson's (D) diversity indices for the different altitude zones on Mt.Biokovo.