ACTA ZOOLOGICA BULGARICA Acta zool. bulg., 54 (3), 2002: 55-74
Spatial Patterns of Terrestrial Small Mammal Communities in Central Western Bulgaria (Mammalia: Insectivora, Rodentia) Teodora V. Minkova*, Vasil V. Popov* Abstract: Capture data of 3215 small mammals belonging to 19 species collected by 200 constantly positioned pitfall traps, exposed during the snow-free seasons of three years (1997 - 1999) in four areas in Central West Bulgaria were examined to obtain information on small mammal habitat associations. Forty environmental variables and (micro)habitat characteristics, considered potentially important for small mammals, were recorded around each trap. The collected data were analyzed by using two complementary multivariate approaches. Hierarchical classification (TWINSPAN) identified eight groups of individual pitfall samples according to their small mammal associations, representing paricular microhabitat types as perceived by the small mammals. The fact that the individual samples within a particular sampling site belong to different types demonstrates the role of microhabitat in shaping the local associations. The ordination analyses (principal component analysis, PCA and redundancy analysis, RDA) demonstrated the role of the regional altitudinal gradient (first ordination axis), the canopy cover (open vs. forested habitats), (second axis), and the humidity related to the local topography (third axis). The local small mammal associations in mountainous sites showed higher species richness, diversity, and abundance compared to those at low altitude. Correlation analyses and comparisons with data from other parts of the country demonstrated the role of humidity in the increase of the abundance and diversity of small mammal associations. Key words: shrews, rodents, microhabitats, environmental gradients, multivariate analyses, diversity.
Introduction In general, the mountainous landscapes along the western border of Bulgaria are still poorly studied in respect to their biodiversity and identified by the Bulgarian Biodiversity Support Program (1994) as a territory requiring special attention and study. No doubt, this region with its variable topography and particular geographic location on the boundary between submediterranean and moderately continental climatic influences in the central part of the Balkan Peninsula represents an important territorial segment of the Bulgarian biota, especially having in mind that it is poorly affected by large scale anthropogenic disturbances. Nevertheless, the ecosystems in the area come under increasing background pressure and are influenced by local activities leading to a substan*Institute of Zoology, Bulgarian Academy of Sciences, 1 Tsar Osvoboditel Blvd, 1000 Sofia, Bulgaria, e-mail:
[email protected]
55
tial loss or degradation of habitat such as timber harvesting (including illegal logging), overgrazing, and recreation building. The landscapes in the area represent a mosaic of vegetational patches reflecting interactions between site characteristics and human activity. Small mammals, being rather sensitive to the effect of numerous environmental factors, are able to respond readily to spatial heterogeneity and can be useful indicators of environmental change and of biodiversity in general. Understanding relationships between small mammals and environmental heterogeneity may therefore help to comprehend implications of management for biodiversity as a whole. In these respects it is important to understand their responses to environmental variables at variable spatial scales - from a microhabitat to a region. The distributional patterns and habitat requirements of small mammals are well known in the northern and central parts of Europe (Banach, 1987; Gaisler, 1983; Andera, 1992; Schropfer, 1990), but few studies concern the southern part of the continent (Loy, Boitani, 1984; Canova, Fasola, 1991; Popov, 2000). We attempt to increase this understanding by exploring the environmental heterogeneity effects on the small mammal local associations and the regional patterns in species richness and abundance. In particular, we use multivariate analyses of pitfall data from four areas within the region to reveal local distribution, microhabitat selection, and structure of local associations of syntopic terrestrial small mammals in a number of sites representing the microhabitat variation within the semi-natural mountainous landscapes of Central Western Bulgaria.
Material and Methods The present study was accomplished in Central Western Bulgaria (42o30’- 42o80’ N/ 22o40’-23o00’ E) on a total area of approximately 1085 km2. Four areas were studied within the region: Milevska Mountain (A), Koniavska Mountain (B), Zemen Gorge (C), and Rui Mountain (D) with an overall altitudinal range from 600 to 1700 m above sea level (a. s. l.). The sampling of small mammals was carried out in 16 sites (A1-A4, B1-B5, C1-C3, D1-D4); their description is presented in Table 1. The sampling in each site was conducted by 12-15 pitfall traps, placed in a line at a distance of about 20 m. Each trap consisted of a plastic cylinder with a diameter of 8 cm and depth of 25 cm, dug in the ground and filled with 5-14% formalin solution up to 1/3 of its volume. A total of 200 pitfall traps with a permanent location were used for a period of three years (from April 1997 to October 1999). Traps were checked every month during the snow-free season which in the area usually lasts from April to October. A total of 3215 individuals of 19 species were captured during the study period. Ten of these species represented 98% of the total catch: Sylvaemus sp. - 530 specimens, Sorex araneus Linnaeus, 1758 - 503, Microtus ex gr. arvalis (Pallas, 1778) - 493, Sorex minutus Linnaeus, 1776 - 339, Clethrionomys glareolus (Schreber, 1780) - 263, Microtus subterraneus (de Selys-Longchamps, 1836) - 232, Crocidura leucodon (Hermann, 1780) - 214, Sylvaemus flavicollis (Melchior, 1834) - 195, Neomys anomalus Cabrera, 1907 - 189, N. fodiens (Pennant, 1771) - 103, Crocidura suaveolens (Pallas, 1811) - 93. The other eight collected species were Muscardinus avellanarius (Linnaeus, 1758) - 24 specimens, Arvicola terrestris (Linnaeus, 1758) -15, Dryomys nitedula (Pallas, 1778) - 7, Sylvaemus sylvaticus (Linnaeus, 1758) - 6, S. mystacinus (Danford, Alston, 1877) - 4, Talpa europaea Linnaeus, 1758 2, Glis glis (Linnaeus, 1766) - 2, and Mus musculus (Linnaeus, 1758) - 1. After being taken out of the pitfalls all small mammals were measured (length of the head and body, length of the tail, length of the hind foot, length of the ear), and then
56
Table 1. Description of the sampling sites. Site A1
Area Milevska mountain Milevska mountain
Altitude above sea level (m) 850
Description edge of the beech forest and short-grass meadow
A3
Milevska mountain
800
young beech forest with closed canopy and poorly developed lower vegetation stratum river-bank deciduous forest with well developed grass and shrub cover
A4
Milevska mountain
900
old oak forest with poorly developed lower vegetation stratum
B1
Konjavska mountain
1450
B2
Konjavska mountain
1400
100-year old beech forest with poorly developed lower vegetation stratum
B3
Konjavska mountain
1400
one year old beech clearing with diverse grass and shrub stratum
B4
Konjavska mountain
1350
B5
Konjavska mountain
1350
C1
Zemen gorge
600
C2
Zemen gorge
600
C3
Zemen gorge
600
D1
Rui mountain
800
mature beech forest with poorly developed lower vegetation stratum
D2
Rui mountain
850
river-bank deciduous forest with well developed grass and shrub cover
D3
Rui mountain
1700
D4
Rui mountain
1650
A2
800
middle mountain damp short-grass meadow
artificially rarefied old beech forest young beech forest with closed canopy and poorly developed lower vegetation stratum dry hornbeam forest with open canopy river-bank deciduous forest with well developed grass and shrub cover dry broad-leaf forest alongside a seasonal brook
high-mountain, damp, short-grass meadow edge of the beech forest and damp meadow
dissected in order to describe the appearance of the uterus, the number and size of embryos - for females, and size of the testicles - for the males. Skulls were cleaned and deposited in the collection of the Institute of Zoology, Bulgarian Academy of Sciences. All skulls were investigated by routine morphometric methods to distinguish the similar species and/or to determine the age. The discrimination between Crocidura leucodon and C. suaveolens meets some difficulties in Bulgaria, where the second species reaches a considerable size. However, the histogram of the height of the procesus coronoideus of the mandible of the available
57
specimens showed a clear bimodality and demonstrated a little overlap between these species. Accordingly, the white toothed shrews with a height of coronoid process of more than 4.6 mm was referred to C. leucodon, and those with a coronoid process height of less then 4.6 mm - to C. suaveolens. The water shrews (Neomys anomalus and N. fodiens) were determined on the basis of a scatter diagram of the length of the hind foot against the length of the tail. The collected specimens formed two well separated clusters. The specimens in each cluster showed well-pronounced differences in the external diagnostic features such as the fur coloration, the development of the tail keel and stiff hairs on hind food, etc. Discrimination between Sylvaemus flavicollis and S. sylvaticus was possible only for adult individuals (age group III and IV of Adamczewska-Andrzejewska, 1967). The exact determination of younger specimens on the basis of morphological criteria was doubtful and they were denoted as Sylvaemus sp. Nevertheless, taking into account that the majority of adult wood mice belonged to S. flavicollis it can be supposed that the prevailing number of undetermined specimens belonged to this species, too. Since a reliable differentiation of Microtus arvalis and M. epiroticus Ondrias, 1966, on the basis of external measurements and skull morphology is still problematic (Gerasimov et al., 1984; Vohralik, 1985; Vohralik, Sofianidou, 1992) we considered the material of the field voles as Microtus ex gr. arvalis. The pooled catch for the three years of the study from each pitfall trap was considered an elementary sample. Thus, the initial data matrix represented the quantitative distribution of small mammal species (rows) across 200 pitfall samples (columns). These data were treated using two complementary multivariate types of analysis: classification and ordination. We used two-way indicator species analysis (TWINSPAN), (Hill, 1979) to provide a classification of the elementary samples. The method results in a double classification of columns and rows in which samples with a similar species composition (and therefore presumably similar environments) were grouped together. The method produces an ordered two-way table allowing interpretation of (micro)site-species relationships. The ordination analyses were done by principal component analysis (PCA) and redundancy analysis (RDA). Usually the correspondence analysis (CA) is considered a more general ordination approach. Here we used PCA, because a preliminary ordination, based on detrended correspondence analysis (DCA), demonstrated relatively short gradients, expressed in standard deviation (SD) units (Ter Braak, 1987): lengths of the first four ordination axes were 2.32, 2.02, 1.57, 0.96 SD. This result presumes a linear response model in which the abundance of any species either increases or decreases along the gradient of latent environmental variables. RDA is a constrained variant of PCA which offers a possibility to incorporate environmental data directly in the ordination. As a result the sample scores are constrained to be linear functions of the measured environmental variables and thus the interpretation of the ordination axes is more straightforward (Ter Braak, 1985, 1987). RDA was used for verification and refinement of the results obtained by the “indirect” methods (TWINSPAN, PCA). PCA and RDA generate ordination diagrams of samples (points) and species (arrows), (PCA) or samples, species, and environmental variables (arrows), (RDA) in the plane of pair-wise sets of axes resulting from the analysis. Each axis explains a percentage of the variance among the samples or species. The dispersion of sample points along a given axis is an expression of an ecological gradient. The direction and length of the species arrows represent the species quantitative change along the gradients expressed by the ordination axes. Similarly, the arrows of environmental variables display the direction of change of a particular variable along the gradients identified by the analysis on the basis of species distribution across samples. In statistical programs
58
used in the analyses (BIODIV, CANOCO, TWINSPAN) and in the resulting diagrams the species names are shown as abbreviations, consisting of three letters of the generic name and three of the species name. Since the interpretation of ordination diagrams representing a scatter plot of the 200 initial samples was difficult, the number of samples was reduced by combining the elementary trap samples from a particular sampling site on the basis of their belonging to a particular typological unit as defined by TWINSPAN (Table 2 - C). The resulting 51 new samples were considered as representing local small mammal associations related to a particular microhabitat type as perceived by small mammals themselves. These samples were designated by the site index (A1-D4) followed by the number of the respective TWINSPAN end group (1 - 8), (Table 2). Before the ordination analyses the new samples were standardized to percentage of the sample total and log (x+1) transformed. The option “downweighting of rare species” was used in order to further minimize the effect of data heteroscedasticity. In RDA, the environmental variables were chosen by a forward selection and the statistical significance of their influence was tested by Monte Carlo permutation test. The Spearman rank correlation coefficient was used to examine the relationships of the diversity indices with the environmental variables and the habitat characteristics. To compare species abundances and diversity indices of local small mammal associations among microhabitat types as defined on the basis of TWINSPAN classification of the initial pitfall samples, we employed Kruskal-Wallis one-way ANOVA. We used analyses of ranked data because various diagnostic tests indicated that the data were not normally distributed. We used rarefaction for independent examination of species richness (Simberloff, 1978; James, Rathbun, 1981). In addition we applied Simpson’s inverse index (N2) to evaluate the species heterogeneity. The rarefaction is a statistical method for estimating the number of species expected in a random sub-sample drawn from a larger sample. Estimates for local associations were adjusted to a sample size of 20 specimens (ES20). The samples with smaller sizes were omitted in the respective analyses. The value of N2 may roughly be interpreted as the equivalent number of species, if they were evenly abundant. For all statistical tests our a priori level of significance was P
