Body size inequality of carabids along an urbanisation gradient

UNCORRECTED PROOF Basic and Applied Ecology ](]]]])]]]—]]] Body size inequality of carabids along an urbanisation gradient Unterschiede der Ko¨rpergro¨ße bei Carabiden entlang eines Urbanisationsgradienten Tibor Magura a,  ,Be´la To´thme´re´sz b ,Ga´bor L. Lo¨vei c a Hortoba´gy National Park Directorate, H-4002 Debrecen, POB. 216, Hungary b Department of Ecology, University of Debrecen, H-4010 Debrecen, POB. 71, Hungary c Department of Integrated Pest Management, Danish Institute of Agricultural Sciences, Flakkebjerg Research Centre, DK-4200 Slagelse, Denmark Received 3 March 2005; accepted 8 August 2005 KEYWORDS Ground beetles; Disturbance; Globenet project; Skewness; Gini coefficient; Lorenz asymmetry coefficient Summary Analysis of size inequality can shed light on coexistence mechanisms and help to interpret patterns in assemblages. We tested several measures for their power to evaluate changes in carabid body size along an urbanisation gradient (city park–suburban area–rural), representing decreasing intensities of human disturbance. Carabids were collected by pitfall traps over two full activity periods in lowland oak forest patches in and near the city of Debrecen, Eastern Hungary. The average value of skewness was largest in the urban areas compared to the suburban and rural ones, indicating that small individuals were more prominent in the urban areas. The Gini coefficient also decreased from urban towards rural areas, suggesting that inequality in body size of the carabid assemblages decreased along the gradient. However, neither of these trends was significant. The Lorenz asymmetry coefficient was significantly higher in rural areas compared to suburban and urban areas indicating that there was a significant difference in inequality and/ or asymmetry of body size across the gradient. This difference was primarily due to more individuals with larger body size in rural area. We suggest that the observed variation in carabid body size along the gradient is related to habitat alteration caused by urbanisation. &2005 Published by Elsevier GmbH on behalf of Gesellschaft fu¨r O ¨kologie. 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 ARTICLE IN PRESS www.elsevier.de/baae 3B2v8:06a=wðDec 5 2003Þ:51c XML:ver:5:0:1BAAE : 50051 Prod:Type:FTP pp:111ðcol:fig::NILÞ ED:Ashim PAGN:anu SCAN:kprashanth 1439-1791/$ - see front matter &2005 Published by Elsevier GmbH on behalf of Gesellschaft fu¨r O ¨kologie. doi:10.1016/j.baae.2005.08.005  Corresponding author. Tel.: +36 52 529920; fax: +36 52 529940. E-mail addresses: [email protected] (T. Magura), [email protected] (B. To´thme´re´sz), gabor[email protected] (G.L. Lo¨vei). UNCORRECTED PROOF Zusammenfassung Die Analyse von Gro¨ßenunterschieden kann die Koexistenzmechanismen beleuchten und helfen, Muster von Ansammlungen zu interpretieren. Wir untersuchten verschiedene Parameter auf ihr Potenzial die Vera¨nderungen der Ko¨rpergro¨ße bei Carabiden entlang eines Urbanisationsgradienten (Stadtpark–Vorstadtgebiet–la¨ndlich) zu erkla¨ren, der eine abnehmende Intensita¨t der anthropogenen Sto¨rung repra¨sentierte. Die Carabiden wurden in Bodenfallen u¨ber zwei vollsta¨ndige aktive Perioden in Tieflandeichenwa¨ldern und in der Na¨he der Stadt Debrecen im o¨stlichen Ungarn gesammelt. Der durchschnittliche Wert der Schiefe war in den sta¨dtischen Gebieten im Vergleich zu den vorsta¨dtischen und la¨ndlichen am gro¨ßten und wies darauf hin, dass kleine Individuen in den sta¨dtischen Gebieten mehr hervortraten. Der GiniKoeffizient verringerte sich ebenfalls von den sta¨dtischen zu den la¨ndlichen Gebieten und la¨ßt vermuten, dass die Unterschiede in der Ko¨rpergro¨ße in den Carabiden-Ansammlungen entlang des Gradienten abnahmen. Keiner dieser Trends war jedoch signifikant. Der Lorenz-Asymmetrie-Koeffizient war in den la¨ndlichen Gebieten im Vergleich zu den vorsta¨dtischen und sta¨dtischen Gebieten signifikant gro¨ßer und wies darauf hin, dass es einen signifikanten Unterschied in den Ko¨rpergro¨ßeunterschieden und/oder in der Asymmetrie der Ko¨rpergro¨ße entlang des Gradienten gab. Dieser Unterschied war vor allem darauf zuru¨ckzufu¨hren, dass es in den la¨ndlichen Bereichen mehr Individuen mit einer gro¨ßeren Ko¨rpergro¨ße gab. Wir vermuten, dass die beobachtete Variation in der Ko¨rpergro¨ße der Carabiden entlang des Gradienten mit der durch die Urbanisation verursachten Vera¨nderung der Habitate verbunden ist. &2005 Published by Elsevier GmbH on behalf of Gesellschaft fu¨r O ¨kologie. Introduction Relationships between body size and the structure of animal assemblages have been the focus of much attention in ecological studies. Body size is correlated with many aspects of life history (reproduction rate, dispersal, development time, etc.) (Peters, 1983). Body size also has a significant impact on ecological interactions, resource use, and more indirectly, the period of activity, habitat suitability, and numerous other parameters (Peters, 1983). A change in body size, either in individual species, or in the size distribution of the species present in a habitat is also a parameter potentially indicating different types of environmental stress (McGeoch, 1998). A number of anthropogenic activities, including farming, forestry and urbanisation, have a significant impact on the environment and create patchworks of modified land types that exhibit similar patterns throughout the world (Poschlod, Bakker, & Kahmen, 2005;Ulrich & Buszko, 2004). Global urbanisation has caused the loss of vast amounts of habitat and caused major modifications of the environmental conditions (Tarvainen, Markkola, & Strommer, 2003). However, little is known on whether or not these changes affect biodiversity in similar ways across the globe (Niemela¨ et al., 2000). In 1998, an international collaborative effort to search for generalisations in urbanisation impacts on biodiversity was initiated. The project, called Globenet (Niemela¨ et al., 2000), examines urban–suburban–rural gradients, using a common methodology and target invertebrate taxon (ground beetles; Fam. Carabidae) (Niemela¨ et al., 2000). This taxon was selected, because carabids are especially useful ecological indicators to study environmental impacts, being sensitive to habitat modifications and environmental changes, abundant and sufficiently variable both taxonomically and ecologically (Lo¨vei & Sunderland, 1996). The results published so far focussed mainly on the changes of carabid assemblage composition along the gradient (Niemela¨ et al., 2002), with some consideration of the effects of urbanisation on body size (Ishitani, Kotze, & Niemela¨, 2003;Magura, To´thme´re´sz, & Molna´r, 2004). Variation in body size has traditionally been described and analysed using the skewness of the size distribution, or other statistics derived from the statistical moments of the distribution (Sokal & Rohlf, 1995). Recently, the focus has shifted toward an emphasis on inequality in size. Several measures of inequality, developed for use in economics (Sen, 1973), have been used to analyse variation in size within assemblages. These measures use the Lorenz curve (Lorenz, 1905), where individuals are ranked by size, and the cumulative proportion of study 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 ARTICLE IN PRESS BAAE : 50051 T. Magura et al.2 UNCORRECTED PROOF objects is plotted against the cumulative proportion of their total size on the y-axis. If all individuals are equal in size, the Lorenz curve is a diagonal line, called the ‘‘line of equality’’ (Fig. 1). Inequality causes the line to run below this line, and the greater the inequality among the study objects, the lower the curve runs below the line of equality. One approach to quantify this is the Gini coefficient (Dixon, Weiner, Mitchell-Olds, & Woodley, 1987;Gini, 1912;Sen, 1973). However, the Gini coefficient is only related to the size (area) and not the shape of the curve. Thus, the Gini coefficient does not contain all the information in the Lorenz curve. Different Lorenz curves can have the same Gini coefficient (Damgaard & Weiner, 2000;Shumway & Koide, 1995;Weiner & Solbrig, 1984). Therefore, Damgaard and Weiner (2000), to characterise the shape of the Lorenz curve, proposed a so-called ‘‘Lorenz asymmetry coefficient’’. This coefficient characterises an important aspect of the shape of a Lorenz curve: it shows which size classes contribute most to the total inequality of the assemblage (Damgaard & Weiner, 2000). The new index was illustrated by an example from plant ecology, but we seek to extend its use to the analysis of size relationships in animal assemblages. In this study, we used pitfall data, collected across an urban–suburban–rural gradient over 2 years, to analyse the body size inequality of ground beetle (Carabidae) assemblages. Using measures describing asymmetry and/or inequality of body size pattern in carabid assemblages, a hypothesis, suggested by Szyszko (1983),Gray (1989) and Blake, Foster, Eyre, and Luff (1994) was tested. According to this ‘‘decreasing body size hypothesis’’, smaller carabids should be found in habitats with higher disturbance levels than in those with lower disturbance. In our case, the hypothesis predicts that the mean carabid body size should decrease from the rural to the urban area. We found that the Lorenz asymmetry index was the most powerful method to detect trends in size along the gradient. Material and methods Characterising the body size distributions The following measures were used to describe the asymmetry and/or inequality of body size pattern in carabid assemblages. (1) Skewness. The asymmetry of a univariate continuous distribution is commonly measured by the classical skewness coefficient (Sokal & Rohlf, 1995), which is defined as g¼Sn i¼1ðxi¯ xÞ3 ns3, where nis the number of individuals, x i is the body size of individuals i,¯ xis the mean body size and sis the standard deviation of body size. A symmetric distribution has zero skewness, i.e. g¼0. An asymmetric distribution with a longer left tail has negative skewness (in our case: large individuals are dominant), while a positive gindicates skewness to the right (smaller individuals are dominant). (2) Medcouple. Since the skewness estimator is based on the first three moments of the data set, it is strongly influenced by the presence of outliers; thus, a robust measure of skewness, the medcouple (Brys, Hubert, & Struyf, 2004) was also used. It has a 25% breakdown value and a bounded influence function. The possible values of medcouple range from 1to1. For notational convenience, the elements of the data set are sorted such that x½1px½2ppx½n. Let med(X n ) denote the median of the data set X n , defined as medðXnÞ¼ ðx½n=2þx½n=2þ1Þ=2ifnis even; x½nþ1=2if nis odd: ( The medcouple is defined as MCn¼med hðx½i;x½jÞ;x½ipmedðXnÞpx½j  , where the kernel function his defined by 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 020406080100 0 20 40 60 80 100 population C population B population A line of equality Cumulative percentage of body size Cumulative p ercenta g e of individuals Figure 1. Lorenz curves of three hypothetical populations. All populations have the same Gini coefficient, but different Lorenz asymmetry coefficients. In the case of the population A, the Lorenz asymmetry coefficient (S)is larger than one (S41), in the case of population B, So1, while size distribution in population Cis symmetric, leading to S¼1. Body size inequality 3 UNCORRECTED PROOF hðx½i;x½jÞ¼ðx½imedðXnÞÞ  ðmedðXnÞx½jÞ x½ix½j for all x½iax½j.Brys et al. (2004) also described a fast algorithm to compute the value of the medcouple. (3) Gini coefficient. A traditional graphical approach to measure inequality in size distribution is the Lorenz curve (Sen, 1973;Weiner & Solbrig, 1984). Individuals are ranked by size and the cumulative proportion of individuals is plotted against the corresponding cumulative proportion of their total size. When all individuals are of the same size, the Lorenz curve is a straight diagonal line, called the line of equality. If there is any inequality in size, the Lorenz curve runs below the line of equality (Fig. 1). The total amount of size inequality can be quantified by the Gini coefficient (Gini, 1912), which is the ratio between the area enclosed by the line of equality and the Lorenz curve, and the total triangular area under the line of equality. The Gini index of aggregation is based on ordered data by increasing body size as follows (Dixon et al., 1987): G¼Pn i¼1ð2in1Þx½i n2¯ x, where nis the number of individuals, x½iis the ordered body size of individuals iand ¯ xis the mean body size. The Gini coefficient calculated by the above equation should be multiplied with n=ðn1Þto obtain an unbiased estimate (Glasser, 1962). The Gini coefficient ranges from a minimum value of zero, if all individuals have the same body size, to a maximum of one in a hypothetical assemblage in which every individual except one has a size of zero. However, it has been demonstrated (Damgaard & Weiner, 2000; Shumway & Koide, 1995;Weiner & Solbrig, 1984) that different Lorenz curves (assemblages with different inequality in size) can have the same Gini coefficient (example on Fig. 1). (4) Lorenz asymmetry coefficient. To complement the above-mentioned Gini coefficient, Damgaard and Weiner (2000) proposed the Lorenz asymmetry coefficient, to quantify the asymmetry of the Lorenz curve. The coefficient (S) can be calculated from the ordered body size data using the following equations (Damgaard & Weiner, 2000): S¼Fð^ xÞþLð^ xÞ¼mþd nþLmþdx0mþ1 Ln , where d¼¯ xx0m x0mþ1x0m and ¯ xis the mean body size, mis the number of individuals with a body size less than ¯ x,L m is the cumulative body size of individuals with a body size less than ¯ x, and L n is the cumulative body size of all individuals. When S¼1, the Lorenz curve of the assemblage is symmetric, while other Svalues represent asymmetric Lorenz curves. When S41, most of the inequality within the assemblage is due to the largest individuals, which disproportionately contributes to the cumulative body size (mass) of the assemblage. When So1, the inequality demonstrated in the assemblage is due primarily to the relatively large number of small individuals (Fig. 1; Damgaard & Weiner, 2000). Study area and sampling methods Ground beetles were studied along an urban–suburban–rural gradient in Debrecen (Eastern Hungary), the second largest city of the country (Magura et al., 2004). The urban, suburban and rural sampling areas were all part of a oncecontinuous forest (Nagyerd+ o Forest Reserve) bordering the city. All areas were situated in continuous patches of old forest (4100 years) dominated by English oak (Quercus robur). The typical, native forest association of the sampling sites was Convallario-Quercetum. The criterion for distinguishing sampling areas (urban, suburban, rural) was the ratio of the built-up area to the natural habitats. The area of the built-up environment and the natural habitats was measured by the ArcView GIS program using an aerial photograph. In the urban area the built-up area exceeded 60%, in the suburban area it was approximately 30%, while in the rural area the built-up area was 0%. The forest fragments in the urban area were parks, where several paths with asphalt surfaces had been created and the shrub layer was strongly thinned. In the suburban area fallen trees were removed. There were occasional, low-intensity forestry management operations in the rural site. Distance between the sampling areas (urban, suburban, rural) was at least 1 km, as prescribed by the general methodology of the Globenet project (Niemela¨ et al., 2000). Four sites, at least 50 m from each other (in order to achieve independence, see Digweed, Currie, Ca´rcamo, & Spence, 1995), were selected within each sampling area. Carabids were collected at each of the four sites of the three sampling areas 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 T. Magura et al.4 UNCORRECTED PROOF using pitfall traps over two full activity periods in 2001 and 2002. Ten traps were placed randomly at least 10 m apart at each site. This resulted in a total of 120 traps scattered along the urban–rural gradient (3 area 4 sites 10 traps). Each pitfall trap was at least 50 m from the nearest forest edge, in order to avoid edge effects (Molna´r, Magura, To´thme´re´sz, & Elek, 2001). Further information on trap design, placement, mode of operation as well as the general description of the collected assemblages are given by Magura et al. (2004). For the present paper, we only used body size, collected from the literature (Hu ˚rka, 1996), and the number of individuals in the catch. For species where minimum and maximum sizes were given, we used the mid-range value (see Table 1). To test for differences in the measures describing asymmetry and/or inequality of body size pattern in carabid assemblages among the three sampling areas, repeated measures analyses of variance (ANOVA) were performed (Sokal & Rohlf, 1995). When the ANOVA revealed a significant difference between the means, LSD (least significant difference) tests were performed for multiple comparisons among means. The analyses were carried out using the R package (R Development Core Team, 2004) and the SPSS-PC program. Results In both years, values of skewness were largest in the urban areas and smallest in the suburban ones indicating that more small individuals were present in the urban areas than in either of the other two (Fig. 2A). The differences, however, were not statistically significant (Table 2). Similar results were obtained with the robust measure of skewness, the medcouple. The values were highest in the urban areas and lowest in the suburban ones (Fig. 2B). These differences were considered significant (Table 2). The Gini coefficient was highest in the urban areas, and decreased towards the rural areas, suggesting that body size inequality of carabid assemblages was largest in the urban parks, and decreased along the urbanisation gradient (Fig. 2C). Here the year treatment interaction was significant, but neither of the component factors was (Table 2). The Lorenz coefficients had values S41 for all situations, indicating the importance of large individuals for the shape of the Lorenz curve. In urban areas, the value was very close to S¼1 in both years. This is a characteristic of a nearly symmetric Lorenz curve. In both years, the rural areas had the highest S values, and the suburban areas had intermediate ones (Fig. 2D). The differences in the Lorenz asymmetry coefficients among the studied areas were significant (Table 2). The value of this coefficient was significantly higher in the rural areas compared to the suburban and urban areas (differences between these two last areas were not statistically significant). Therefore, the significant difference in the shape of the Lorenz curves was caused primarily by a higher number of individuals with larger body size in the rural area vs. the other two areas under higher degree of urbanisation. Discussion Analysing inequality has a longer history in plant than animal studies. To describe inequality in plant size, several studies (Creed, Kain, & Norton, 1998; Ditommaso & Watson, 1997;Zammit & Zedler, 1993) used skewness derived from the statistical moments of the size distribution. Several other papers used the Gini coefficient to measure inequality in plant size or biomass (Hanley & Groves, 2002;He, Ma, Brown, & Lynch, 2005;Leiss &Mu¨ller-Scha¨rer, 2001;Ramstad & Hestmark, 2001; Shumway & Koide, 1995;Wilson & Gurevitch, 1995). For plants, Damgaard and Weiner (2000) calculated the Lorenz asymmetry coefficient for data from Shumway and Koide (1995) to interpret the effect of mycorrhizae and plant density on the number of capsules produced by Abutilon theophrasti (Fam. Malvaceae) individuals. They were able to show that the reported inequality in the number of capsules when the plants contained mycorrhizae was caused by the increased importance of individuals with high capsule production (Damgaard & Weiner, 2000). This, however, remains the only example of using the proposed index. Skewness is the only method used to evaluate inequality and/or asymmetry in size distribution of animal populations or assemblages (Gomez & Espadaler, 2000;Gregory, 2000;Knouft, 2004; Kozlowski & Gawelczyk, 2002;Novotny & Kindlmann, 1996;Poulin & Morand, 1997). In the present paper, we extended the range of methods applied to analyse the inequality of animal body size distribution using two other parameters (the Gini coefficient and the Lorenz asymmetry coefficient) and showed that the Lorenz asymmetry coefficient, S, is a powerful method for studies describing and interpreting variations in body size. Published studies in the international Globenet project characterised changes in the carabid body size distribution along the urban–suburban–rural 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 Body size inequality 5 UNCORRECTED PROOF 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 Table 1. A list of the carabid species, their body size, and the number of individuals collected in the urban, suburban and rural sampling areas during the two study years near Debrecen, Eastern Hungary Species Body size (mm) Urban Suburban Rural Agonum lugens 9.0 2 0 0 Amara anthobia 6.2 10 0 0 Amara communis 6.5 9 0 0 Amara consularis 8.3 0 1 2 Amara convexior 7.7 113 26 47 Amara familiaris 6.3 16 5 4 Amara lucida 5.6 1 0 0 Amara ovata 8.6 10 0 0 Amara saphyrea 8.8 10 16 39 Amara similata 8.6 2 2 2 Anchomenus dorsalis 6.6 1 0 0 Anisodactylus nemorivagus 8.9 52 0 0 Anisodactylus signatus 11.8 1 1 0 Asaphidion flavipes 4.3 2 0 0 Badister bullatus 5.2 11 1 0 Badister lacertosus 6.3 5 15 1 Badister meridionalis 6.7 8 2 0 Bembidion lampros 3.4 38 0 3 Calathus erratus 9.5 2 0 1 Calathus fuscipes 11.1 26 0 3 Calathus melanocephalus 7.1 1 0 0 Calosoma inquisitor 20.0 0 0 10 Carabus convexus 17.0 1 107 124 Carabus granulatus 19.0 6 1 6 Carabus ullrichi 27.0 1 0 0 Carabus violaceus 28.0 75 78 237 Clivina fossor 5.9 3 0 0 Diachromus germanus 8.4 1 0 0 Harpalus distinguendus 9.5 0 1 0 Harpalus latus 9.1 14 1 24 Harpalus luteicornis 7.1 5 20 1 Harpalus tardus 9.4 104 86 69 Harpalus xanthopus winkleri 7.1 21 10 3 Leistus ferrugineus 6.8 0 0 1 Licinus depressus 10.4 7 5 1 Notiophilus biguttatus 4.9 2 0 0 Notiophilus palustris 5.1 5 2 6 Notiophilus rufipes 5.3 38 14 7 Ophonus nitidulus 9.6 1 2 42 Ophonus schaubergerianus 8.8 0 0 1 Oxypselaphus obscurus 5.5 0 1 1 Panagaeus bipustulatus 7.2 7 5 0 Platyderus rufus 6.3 76 41 79 Poecilus cupreus 11.8 3 0 0 Poecilus versicolor 10.5 1 0 0 Pseudoophonus griseus 10.1 2 0 1 Pseudoophonus rufipes 13.1 10 26 19 Pterostichus anthracinus 11.2 5 3 0 Pterostichus macer 12.9 1 0 0 Pterostichus melanarius 15.7 58 3 1 Pterostichus melas 14.9 2 0 3 Pterostichus minor 7.6 0 1 0 Pterostichus niger 18.4 22 15 23 Pterostichus oblongopunctatus 11.5 117 454 1505 Pterostichus ovoideus 7.1 1 0 0 Pterostichus strenuus 6.0 27 52 11 T. Magura et al.6 UNCORRECTED PROOF gradient using either the distribution among different, arbitrary size classes (Alaruikka, Kotze, Matveinen, & Niemela¨, 2002;Ishitani et al., 2003)or the mean body size of the species weighted by their respective abundance (Gaublomme, Dhuyvetter, Verdyck, & Desender, 2005;Magura et al., 2004; Niemela¨ et al., 2002). In Finland, Alaruikka et al. (2002) investigating the changes of carabid body size across an urbanisation gradient concluded that mediumto large-sized carabid individuals were more likely to be collected in the rural sites than in urban forest fragments. In Japan, there are no large and only few medium-sized specialist species in the urban environment, while many specimens of medium-sized and some large-sized specialist species occur in the suburban and rural sites (Ishitani et al., 2003). Mean carabid body size changed significantly from small values in the urban area to larger ones in both suburban and rural areas in Bulgaria (Niemela¨ et al., 2002), Hungary (Magura et al., 2004) and Belgium (Gaublomme et al., 2005). There was a marginally significant change in the same direction along the same gradient in Finland, but none in Canada (Niemela¨ et al., 2002). However, not only body size of the carabid assemblages may change across an urbanisation gradient; there could be changes among different populations of the same species. The body size of Carabus nemoralis decreased significantly from the rural surroundings of Hamburg, Germany, towards the city centre (Weller & Ganzhorn, 2004). The present study, using a more sophisticated method (the Lorenz asymmetry coefficient), not only proved the existence of a significant change in inequality of carabid body size across the urban–- suburban–rural gradient, but indicated that this difference was primarily due to an increase in the contribution of individuals with larger body size in the rural area. The mean body size of ground beetles also increased (Magura et al., 2004), but 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 Table 1. (continued ) Species Body size (mm) Urban Suburban Rural Stomis pumicatus 7.5 1 15 27 Synuchus vivalis 7.2 7 6 146 Trechus quadristriatus 3.5 0 1 1 Species sequence is alphabetical. Urban Suburban Rural Urban Suburban Rural Urban Suburban Rural Urban Suburban Rural 0 1 2 3 (D) (C) (B) (A) Skewness -0.2 0.0 0.2 0.4 Robust skewness 2001 2002 0.0 0.1 0.2 0.3 Gini coefficient 0.0 0.5 1.0 1.5 Lorenz asymmetry coefficient Figure 2. Average values (7S.E.) of the skewness (A), the robust skewness (B), the Gini coefficient (C) and the Lorenz asymmetry coefficient (D) for the urban, suburban and rural carabid assemblages in the two study years. Body size inequality 7 UNCORRECTED PROOF this change can result from a decrease in the importance of small species, from the increase in medium-sized or large species, or a combination of these. By evaluating the mean body size, we cannot distinguish among these possibilities. The Lorenz asymmetry coefficient allowed us to demonstrate which of these theoretical possibilities was responsible for the observed effect. The larger carabid body size in the less disturbed area (rural area) and the smaller body size in the moderately or highly disturbed areas (suburban and urban areas) could be explained by the hypothesis postulated by Szyszko (1983),Gray (1989) and Blake et al. (1994). Szyszko (1983), studying the regeneration of pine plantations after clear-cutting in Poland, suggested and used the mean individual biomass (MIB) index. This index is simply calculated as the ratio of the total fresh body mass of the catch in a trap, divided by the number of carabid individuals caught. Szyszko (1983) showed that as regeneration in the plantation proceeds, the average value of the MIB index also increases. Mean body size is positively related to body mass, and thus the conclusion is that the mean body size in carabid assemblages will also increase. Gray (1989) hypothesised that the mean body size of species should decrease from undisturbed towards disturbed habitats. Carabid assemblages of differently managed grasslands gave support to this hypothesis (Blake et al., 1994). Highly disturbed areas support carabid 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 Table 2. The results of repeated-measures ANOVA for the values describing asymmetry and/or inequality of body size pattern in carabid assemblages Source SS df MS FpLSD test Skewness Tests of within-subjects contrasts Year 0.001 1 0.001 0.008 0.931 Year Area 0.306 2 0.153 0.890 0.444 Error 1.547 9 0.172 Tests of between-subjects effects Area 2.473 2 1.237 1.235 0.336 Error 9.011 9 1.001 Robust skewness Tests of within-subjects contrasts Year 0.058 1 0.058 1.099 0.322 Year Area 0.020 2 0.010 0.192 0.828 Error 0.479 9 0.053 Tests of between-subjects effects Area 0.424 2 0.212 3.868 0.061 U4S Error 0.493 9 0.055 Gini coefficient Tests of within-subjects contrasts Year 0.0001 1 0.0001 0.142 0.715 Year Area 0.008 2 0.004 5.073 0.033 Error 0.007 9 0.0007 Tests of between-subjects effects Area 0.014 2 0.007 1.974 0.195 Error 0.031 9 0.003 Lorenz asymmetry coefficient Tests of within-subjects contrasts Year 0.003 1 0.003 0.044 0.839 Year Area 0.058 2 0.029 0.469 0.640 Error 0.558 9 0.062 Tests of between-subjects effects Area 0.932 2 0.466 7.315 0.013 U¼SoR Error 0.573 9 0.064 Year ¼the effect of study year (2001 and 2002), Area ¼the urban, suburban and rural sampling areas. Results of the LSD test indicate which area(s) differ(s) significantly (po0:05) from the others; for example U¼SoRindicates that the measured value was significantly higher in the rural area than in the urban and suburban area (these two areas, however, were not different). T. Magura et al.8 UNCORRECTED PROOF assemblages with species of smaller average body size than do less disturbed sites (Blake et al., 1994; Grandchamp, Niemela¨, & Kotze, 2000;Holliday, 1991;Magura, Elek, & To´thme´re´sz, 2002;Ribera, Dole´dec, Downie, & Foster, 2001;S ˇustek, 1987). The causes of this can be manifold. Carabids have ground-living larvae that are weakly chitinised, limited in mobility, and thus more sensitive to changing conditions than adults (Lo¨vei & Sunderland, 1996). Disturbance will frequently create unfavourable conditions for ground beetle adults as well as larvae, when their densities decrease (Thorbek & Bilde, 2004) and species may become locally extinct. Small-sized carabid species may suffer less mortality during such disturbance events. Their densities are also usually higher than that of large-sized species (Luff, 2002), so they have a lower probability of local extinction. Small species are more often winged than are large-sized species (Ribera et al., 2001). Consequently, small species are more vagile than large species and can colonise disturbed and unstable areas more easily (Thiele, 1977). Smaller species may also need fewer resources and/or may develop faster than large species (Peters, 1983). In carabids, large species have longer larval periods, making them more vulnerable to disturbance events (Kotze & O’Hara, 2003). Small species can use the small ‘‘windows of suitability’’ to survive in the disturbed habitat. Lo¨vei and Sunderland (1996) also call attention to the importance of larvae to explain trends in adults. Along the studied urbanisation gradient, the degree of disturbance is higher in the urban (paved paths, thinned shrub layer, intensive landscape management) and in the suburban area (management of moderate intensity, e.g. fallen trees are removed) than in the rural area (rare occasions of intervention, low intensity management). Disturbance caused by urbanisation appears to eliminate favourable microsites for forest species with larger body size and create altered, relatively homogeneous micro-habitats invaded by small-sized species capable of flying. All these habitat alterations accompanied by urbanisation contributed to the observed variation in carabid body size across the urban–suburban–rural gradient. Using the Lorenz asymmetry coefficient (Damgaard & Weiner, 2000), we were able to more completely analyse the size distributions of the ground beetle assemblages along the urbanisation gradient. This index has proven to be more powerful than more traditional methods such as skewness, robust skewness (medcouple), or the Gini coefficient. The biological interpretation of the index is not problematic, and we suggest that it is a useful tool for future studies of size/biomass distribution in animal assemblages. Acknowledgements The Globenet study in Hungary was supported by the Hortoba´gy National Park Directorate. T. Magura is a Bolyai Research Fellow of the Hungarian Academy of Sciences. We are especially grateful to Tivadar Molna´r for field and laboratory assistance. We thank two anonymous reviewers for their helpful comments on the manuscript. References Alaruikka, D. M., Kotze, D. J., Matveinen, K., & Niemela¨, J. (2002). Carabid and spider assemblages along an urban to rural gradient in Southern Finland. Journal of Insect Conservation,6, 195–206. Blake, S., Foster, G. N., Eyre, M. D., & Luff, M. L. (1994). Effects of habitat type and grassland management practices on the body size distribution of carabid beetles. Pedobiologia,38, 502–512. Brys, G., Hubert, A., & Struyf, A. (2004). A robust measure of skewness. Journal of Computational & Graphical Statistics,13, 996–1017. Creed, J. C., Kain, J. M., & Norton, T. A. (1998). An experimental evaluation of density and plant size in two large brown seaweeds. Journal of Phycology,34, 39–52. Damgaard, C., & Weiner, J. (2000). Describing inequality in plant size or fecundity. Ecology,81, 1139–1142. Digweed, S. C., Currie, C. R., Ca´rcamo, H. A., & Spence, J. R. (1995). Digging out the digging-in effect of pitfall traps: influences of depletion and disturbance on catches of ground beetles (Coleoptera: Carabidae). Pedobiologia,39, 561–576. Ditommaso, A., & Watson, A. K. (1997). Effect of the fungal pathogen, Colletotrichum coccodes,onAbutilon theophrasti height hierarchy development. Journal of Applied Ecology,34, 518–529. Dixon, P. M., Weiner, J., Mitchell-Olds, T., & Woodley, R. (1987). Bootstrapping the Gini coefficient of inequality. Ecology,6, 1548–1551. Gaublomme, E., Dhuyvetter, H., Verdyck, P., & Desender, K. (2005). Effects of urbanisation on carabid beetles in old beech forests. In: Lo¨vei, G. L., & Toft, S. (Eds.), European carabidology 2003. Proceedings of the 11th European carabidologists’ meeting, DIAS Report, Vol. 114, pp. 111–123. Gini, C. (1912). Variabilita e mutabilita. In E. Pizzetti, & T. Salvemini (Eds.), Memorie di metodologica statistica (pp. 211–382). Rome, Italy: Eredi Virgilio Veschi. Glasser, G. J. (1962). Variance formulas for the mean difference and coefficient of concentration. Journal of the American Statistical Association,57, 648–654. 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 ARTICLE IN PRESS BAAE : 50051 Body size inequality 9