Introduction

Habitat fragmentation is frequently caused by human activities such as agriculture, rural development, and urbanization; and forest edges tend to become more abundant in human-dominated areas. In fragmented habitats, habitat structure and microclimate are often affected by edge creation (see Harper et al. 2005; Laurance 2008), which can alter species composition. Therefore, understanding the edge effect in fragmented environments is important for biodiversity conservation. Indeed, the negative effects of edge creation on habitat specialist species have become apparent (Harper et al. 2005). For example, habitat fragmentation often negatively affects threatened bird species (Watson et al. 2004), and large-scale edge effects can lead to the local extinction of beetle species that are restricted to habitat interiors (Ewers and Didham 2008).

In Korea, humans have begun to use forests and adjacent areas more intensively in recent decades. During the Korean War and earlier, most of the primary forests in Korea were severely damaged, resulting in reduced growing stocks of 5.6 m3/ha in 1952. After this period, forests were regenerated through extensive reforestation, with growing stocks restored to 126 m3/ha in 2010 (Lee 2012). Though reforestation has been successful, rapid industrialization and urbanization have prompted a loss of biodiversity through the transformation of natural habitats (i.e. forests and grasslands) into agricultural and urban environments. From the 1960s to 2014, for example, forest in Korea decreased from 68.6 to 63.9 % of the total land area, with 27.8 % of total land area being used for agriculture and urbanization by the end of that period (Ministry of Land, Transport and Maritime Affairs 2015). This move to maximize land use for urban and agricultural development is still in progress. Disturbances by human activities, such as urbanization and agricultural development, are major causes of biodiversity loss in Korea (Ministry of Environment 2012), and it has become important to examine the effect of human activities on biodiversity at forest edge for planning conservation and management of biodiversity because forest edges are being increased worldwide as well as in Korea.

Ground beetles are useful subjects for comparative ecological studies because they can be caught easily with pitfall traps. Ground beetles constitute a taxonomically and ecologically diverse group; they are well-documented and known to closely reflect environmental conditions. Thus, ground beetles are good indicators of environmental change (Thiele 1977; Lövei and Sunderland 1996; Rainio and Niemelä 2003). In fragmented landscapes, the effects of area and edge can significantly influence ecological processes, but there is little available data that can be used to understand the interaction between edge and area effects for ground beetles. Exceptions are the works produced by Barbosa and Marquet (2002), Ewers et al. (2007), and Soga et al. (2013) in urban landscapes. However, many studies have investigated the response of ground beetles to forest edges in terms of species richness and composition (e.g. Heliölä et al. 2001; Magura 2002; Koivula et al. 2004; Yu et al. 2007; Ewers and Didham 2008; Roume et al. 2011; Tóthmérész et al. 2014; Lacasella et al. 2015; Ohwaki et al. 2015). Most of these studies were conducted in forest–clear cuts or forest–grasslands, and few studies (Koivula et al. 2004; Ohwaki et al. 2015) have been conducted on the forest–farmland edge in agro-forest landscapes. Further, most studies on responses of ground beetles to forest edges have focused on the forest edge itself, whereas the spatial distribution of ground beetles across forest–farm edges in accordance with landscape structure is still poorly understood. In temperate regions, agro-forest landscapes tend to be mosaic landscapes comprising secondary forests and various types of farmland. Because of the relative complexity of agro-forest landscapes, it is difficult to predict how landscape structures will affect biodiversity across forest–farmland edges. One putative effect is that the change from natural grasslands to farmlands among forest patches may restrict the dispersal of many forest specialists because altered soil surfaces in farmland habitats may operate as a barrier for many forest specialists (Koivula et al. 2004). Edge creation is a major factor that can cause decreases in the diversity of arthropods, especially forest specialists (e.g. Ewers and Didham 2008; Halaj et al. 2008). However, forest edges have been also regarded as important habitats for biodiversity conservation because forest–farm edges have the potential to provide temporary refuge and overwintering sites, enabling the dispersal and re-colonization of many predatory invertebrates, including ground beetles (Magura et al. 2001; Molnár et al. 2001; Magura 2002; Yu et al. 2007; Roume et al. 2011; Ohwaki et al. 2015). Thus, studies on how edge effects influence the biodiversity of ground beetles in diverse habitat types and landscapes can capture the impact of edges on the overall biodiversity of habitats with varying spatial complexity.

Therefore, we investigated patterns of species richness, abundance, and species composition of ground beetles across forest–farm transects in two different agro-forest landscapes: one fragmented and developing landscape and one relatively well-protected landscape. In addition, we examined the habitat preferences of each ground beetle species to enable further ecological studies based on their distribution patterns.

Materials and methods

Study landscape

This study was conducted at two sites, Hwaseong and Heongseong, in Korea. Hwaseong is located in the western part of Korea, is highly heterogeneous, and is composed of fragmented landscapes (Fig. 1a). Approximately 36.9 % (25,344.9 ha) of the land area is forest, while 36.1 % (24,854.7 ha) and 14.4 % (9913.0 ha) are farmlands (rice fields, upland farms, and orchards) and urban areas (developed areas, roads, and railways), respectively (Hwaseong Statistical Year Book, http://www.hscity.go.kr). In addition, forest and farmland in Hwaseong have consistently decreased owing to increases in the extent of developed areas and roads.

Fig. 1
figure 1

Locations of 14 and 8 collection sites in Hwaseong (a) and Hoengseong (c), respectively, and sampling designs for ground beetle collection in Hwaseong (b) and Hoengseong (d). Shaded area indicates forest area. Abbreviations of sampling sites are listed in Table 1

Full size image

Hoengseong, located in the eastern part of Korea, is a better preserved agro-forest landscape than Hwaseong (Fig. 1c). Approximately 77.0 % (76,806.1 ha) of the land area is forest, while 15.2 % (15,196.6 ha) and 1.3 % (1286.1 ha) are farmlands (rice fields, uplands, and orchards) and developed areas, respectively (Hoengseong Statistical Year Book, http://www.hsg.go.kr).

Ground beetle sampling and identification

For ground beetle collection, pitfall traps were used. Pitfall trapping is a standard sampling method for comparing the abundance or community structure of ground beetles (Niemelä 1996; Koivula et al. 2003). The plastic pitfall traps (300-ml volume, 75-mm diameter, 120-mm depth) were un-baited, containing preservatives (150 ml, 95 % ethyl-alcohol:95 % ethylene–glycol = 1:1) as killing-preserving solution. A plastic roof was placed 3 cm above each trap to prevent the inflow of rainfall and litter.

In Hwaseong, nine and five forests were selected in 2011 and 2012, respectively, and their interiors, edges and surroundings were studied (Fig. 1a; Table 1). Seven of the forest study sites were conifer forests dominated by Pinus densiflora Siebold et Zucc. and Pinus rigida Mill., while others were mixed forests dominated by oaks (Quercus acutissima Carruthers), Japanese chestnut (Castanea crenata Siebold & Zucc.), and Pinus species. Patch sizes of all forests ranged from 4.6 to 518.5 ha except for site M1 (1548.4 ha) and all forests were regenerating forests or plantations (30–50 years old). The areas surroundings the forests were generally human-managed areas, such as agricultural land, rice field levees, and cemetery turf.

Table 1 Characteristics of each study site
Full size table

The minimum diameter of small forest patches in Hwaseong was about 40 m, with most small patches being highly thinned or reduced from their previous sizes. For this reason, true forest edge–interior distance in Hwaseong was 20 m, and the total length of the gradient across the forest surrounding to the interior was 40 m. Therefore, we divided this area into three separate zones according to their distance from the forest edge; we set three traps at the edge (1 m inward from the forest boundary), three traps in the forest interior (20 m inward from the forest boundary), and three traps in the area surrounding the forest (20 m outward from the forest boundary) (Fig. 1b). In total, nine pitfall traps were placed within each study site and a total of 126 pitfall traps were used in 14 study sites in Hwaseong. Pitfall traps were replaced every month from early June to early October in 2011 and late June to late October in 2012.

Like Hwaseong, Hoengseong also contains secondary forests (40 years old) (Table 1). Four study sites (PF1, PF2, PQF1, and PQF2) were dominated by coniferous tree species (mainly P. densiflora), but the forest edges at PQF1 and PQF2 were mixed with oaks (Quercus spp.). An additional four sites (LQF1, LQF2, QF1, and QF2) were mixed forests dominated by oaks and two of the transects at LQF1 and LQF2 were mixed Larix kaempferi Carriere and oaks.

Unlike the forest in Hwaseong, the forest in Hoengseong is continuously connected and the landscape is well preserved (Fig. 1c). Because the forest is continuous, a set of two pitfall traps was installed along the transect (Fig. 1d). Each transect began in the forest interior 80 m away from the forest edge and extended 80 m into the surrounding area (+80, +40, +20, +1, −20, −40, −80 m). Surrounding habitats included agricultural lands, rice field levees, unmanaged areas, cemetery turf, and a riverbank. Eight transects were set at least 300 m away from each other to ensure adequate spatial independence for statistical analysis of pitfall samples. Overall, 14 pitfall traps were placed in each study site, with a total of 112 pitfall traps used in Hoengseong. Pitfall traps were replaced every month from early May to early October in 2012.

Forest information (forest patch sizes, dominant conifer species, and forest ages) in two study landscapes was confirmed by the forest geographic information service system in Korea (Korea Forest Service 2014).

Collected ground beetles were brought to the laboratory, dried, mounted, and identified to species level under a dissecting microscope (63×). Identification was performed according to Habu (1967, 1973, 1978, 1987), Kwon and Lee (1984), and Park and Paik (2001), and nomenclature was confirmed according to the lists of Korean Carabidae by Park and Paik (2001) and Park (2004).

Data analysis

Species richness was measured by the total number of species collected during the sampling period, and abundance was measured by the total number of individuals collected in a set of traps for each study site. In addition, we conducted two-way analysis of variance (ANOVA) using the generalized linear model to determine the effects of forest type, edge orientation, and distance from the forest edge (+80, +40, +20, +1, −20, −40, −80 m) on ground beetle abundance and species richness. Because ground beetle catches may be influenced by the orientation of the edge (Ries et al. 2004), we classified edges into northern (i.e. forest edge facing from northwest to northeast) and southern (i.e. forest edge facing from southwest to southeast) (Koivula et al. 2004). Data on abundance and species richness analyzed with two-way ANOVA were log(x + 1) transformed for normalization. In addition, the species richness of ground beetles was estimated with individual-based rarefaction curves for each habitat type. Rarefaction curves are based on random re-sampling of the pool of captured individuals, and are used to estimate expected richness in lower sample sizes (Gotelli and Colwell 2001). Rarefaction methods allow for meaningful standardization and comparison of datasets (Gotelli and Colwell 2001). We pooled the data (abundance and species richness) of collected ground beetles according to habitat type (forest interior, edge, and surrounding) in each study landscape. Then, the estimated species richness was compared among habitat types.

To summarize and compare species composition among different habitat types and distance from the forest edge based on square-root transformed data, a matrix of Bray–Curtis dissimilarity values was created (Clarke and Warwick 2001). Similarities among study sites from both Hwaseong and Hoengseong were graphically represented by non-metric multidimensional scaling (NMDS) ordination. NMDS was chosen because it performs well with ecological data that do not meet the assumption of normality (McCune and Grace 2002) and, thus, it is the most generally effective ordination and classification method for ecological community data (Jongman et al. 1995; McCune and Grace 2002). In NMDS, stress represents distortion between real data point positions in a graph. Lower stress represents less distortion from the real data point positions and thus indicates that the graph more accurately represents dissimilarities in species composition.

The indicator value (IndVal) approach was employed to determine indicator species that characterize the habitats (Dufrêne and Legendre 1997). The flexible IndVal was independent of other species’ relative abundance, and there was no need to use pseudo-species. The IndVal is at maximum (100 %) when all individuals of a species are found in a single group of sites and when the species occurs in all sites for that group. Abundance and occurrence stability indexes for species were determined to be important and were included for analysis. Furthermore, it is possible to check the statistical significance of the relationship between species occurrence/abundance and groups of sites (De Caceres 2013).

Two-way ANOVA with Tukey’s post hoc test, rarefaction curves, NMDS based on Bray–Curtis dissimilarity indices, and IndVal were conducted using the R version 3.0.2 (R Core Team 2013).

Results

In Hwaseong, a total of 5276 ground beetles belonging to 42 species were collected from 14 study sites (2125 individuals of 18 species in the forest interior, 2208 individuals of 15 species in the forest edge, and 943 individuals of 38 species in the forest surroundings) (Table S1). Four species, Chlaenius naeviger Morawitz, Synuchus arcuaticollis Motschulsky, Synuchus cycloderus (Bates), and Synuchus nitidus (Motschulsky), accounted for more than 81.3 % of individuals. In Hoengseong, a total of 3741 ground beetles belonging to 61 species were collected from eight study sites (2160 individuals of 21 species in the forest interior, 810 individuals of 25 species in the forest edge, and 771 individuals of 52 species in the forest surroundings) (Table S1). Seven species, C. naeviger, Coptolabrus jankowskii jankowskii (Oberthur), Coptolabrus smaragdinus branickii Taczanowski, Pheropsophus jessoensis Morawitz, S. arcuaticollis, S. cycloderus, and S. nitidus, accounted for more than 80.7 % of individuals.

Individual-based rarefaction curves for all species from Hwaseong and Hoengseong showed that species richness was generally higher in the surrounding habitat than in the forest interior or edge (Fig. 2). In Hwaseong, the species richness of the forest edge was similar to that of the forest interior (Fig. 2a), while in Hoengseong forest edge species richness was between those of the forest interior and surroundings (Fig. 2b).

Fig. 2
figure 2

Individual-based rarefaction curves for ground beetle catches from the forest interior (+80 to +20 m), edge (+1 m) and surroundings (−20 to −80 m) in Hwaseong (a) and Hoengseong (b). Data points indicate average species numbers estimated from the total number of individuals

Full size image

When comparing among forest type, edge orientation, and distance from the forest edge, we found distance from the forest edge to be the most significant factor for the abundance and species richness of ground beetles (Table 2). In both Hwaseong and Hoengseong, the abundance of ground beetles and forest specialists was significantly higher in the forest interior and edge than in the farmland, while that of open-habitat species was higher in the farmland. Similar patterns were found in the species richness of forest specialists and open-habitat species. However, the forest type and edge orientation were not related to the species richness and abundance of forest specialists.

Table 2 Two-way ANOVA results for abundance and species richness of all ground beetles and habitat-associated groups (forest specialists and open-habit species)
Full size table

Combined data from Hwaseong and Hoengseong were used for NMDS based on the Bray–Curtis dissimilarity coefficient. Overall, the species composition of ground beetles in the forest edge was more similar to that in the forest interior than to that in the forest surroundings (Fig. 3). However, a difference between Hwaseong and Hoengseong was found according to axis 2, not only in the forest interior (analysis of similarities, Global R = 0.237, P = 0.001) and edge (Global R = 0.378, P = 0.002), but also in the forest surroundings (Global R = 0.118, P = 0.024) (Fig. 3a). The species composition of ground beetles was not distinguished by forest types (i.e. pine-dominated, oak-dominated or mixed forest) or by surrounding habitat types (i.e. agricultural land, levees, turf, river bank, or unmanaged areas) (Fig. 3b).

Fig. 3
figure 3

Non-metric multidimensional scaling using Bray–Curtis dissimilarity indices, based on square-root transformed data of ground beetle assemblages in both Hwaseong and Hoengseong according to distance from the forest edge (a) and habitat type (b)

Full size image

Three characteristic groups of ground beetle species were detected by IndVal (Table 3): forest specialists, which were numerous in most of the forest interior and edge areas; edge-associated species, which were most abundant in the forest edge and surroundings, and open-habitat species, which were most abundant in open habitats such as grasslands and farmland. Some species, such as C. naeviger and P. jessoensis, were not included in these groups by IndVal, but they were abundant in the forest edge as well as the forest interior or surroundings (Fig. 4a, c). One species—C. smaragdinus branickii—was mainly caught in forest interiors and edges in Hoengseong, but mostly found in surrounding habitats in Hwaseong (Fig. 4o).

Table 3 Two-way indicator species analysis results showing species indicator values and the number of individual ground beetles by habitat type in two different agro-forest landscapes
Full size table
Fig. 4
figure 4

Spatial distribution of 15 abundant species along the forest interior (+80 to +20 m)–edge (+1 m)–farm (−20 to −80 m) transects in Hoengseong and Hwaseong

Full size image

Discussion

Our results indicate that ground beetle assemblages were primarily influenced by the forest edge according to overall landscape structure (i.e. fragmented or continuous forest). The species richness and composition of ground beetle assemblages in the forest edges of Hwaseong, a fragmented landscape, were similar to those of the forest interior. In Hoengseong, a relatively well-protected landscape, forest edge species richness was intermediate between that of the forest interior and that of the forest surroundings because of the varying species distribution along forest–farm transects. The proportion of forest, arable, and urbanized land area in each landscape may be significant factors affecting ground beetle assemblages at forest edges, especially for forest specialist species.

Ground beetle assemblages in forest–farm edges

For ground beetles, higher species richness has generally been found in forest edges and surroundings than in forest interiors along forest–field transects (e.g. Niemelä and Halme 1992; Halme and Niemelä 1993; Bedford and Usher 1994; Levesque and Levesque 1994), and forest–grassland transects (e.g. Molnár et al. 2001; Magura 2002; Tóthmérész et al. 2014; Lacasella et al. 2015). In our study, species richness and composition of ground beetles were examined in terms of the edge effect along forest–farmland transects, with patterns of species richness found to differ between the two agro-forest landscapes (fragmented and preserved). In Hoengseong, the species richness of the forest edge was between that in the forest interior and surroundings, while no difference was found between the forest edge and interior in Hwaseong. This difference may be a result of the different degrees of habitat fragmentation and urbanization because the species richness of forest specialist ground beetles generally decreases with decreasing patch size (Magura et al. 2010); further, open-habitat species are often affected by land-use patterns (Vanbergen et al. 2005). Estimated species richness in the forest edge was higher in Hoengseong mainly because of the many open-habitat species caught there (Fig. 2b). In fact, habitats in Hoengseong appear to be more stable than those in Hwaseong in terms of the forest continuity and lower rate of urbanization. From this finding, we project that biodiversity at forest edges would decrease if natural open-habitats and gradual edges were degraded by urbanization and fragmentation. Habitat heterogeneity is an essential structure for animal species diversity (Tews et al. 2004), and the dual presence of open habitats and gradual edges should increase habitat heterogeneity in agro-forest landscapes. Notably, the biodiversity of ground beetles in gradual edges (i.e. shrubby forest edges with much ground vegetation and a shrub layer) was found to be higher than in abrupt edges (i.e. unstable edges fragmented by farmlands or roads); and species composition in these different edge types also differed (e.g. Do and Joo 2015).

In both Hwaseong and Hoengseong, the species composition of ground beetles was very similar between the forest edge and interior compared with that of the surrounding habitats. This result agrees with a study by Koivula et al. (2004). For ground beetles, the dispersal of many forest specialists in a heavily fragmented landscape may be restricted to within the forest. This is because canopy cover is known as a major determining factor in the spatial distribution of ground beetles (Koivula et al. 2004; Vanbergen et al. 2005). Although several forest specialists (e.g. Calosoma spp., Coptolabrus spp., and Synuchus spp.) in our study were found in surrounding habitats, their distribution was mostly restricted to the forest. This may be due, in part, to potential differences in the soil surface properties between farmlands and forests (Koivula et al. 2004). Thus, to minimize the adverse effects of farmlands on ground beetle dispersal, forest edges in agro-forest landscape should be protected by expanding field margins to provide additional, diverse microhabitats.

Many forest specialists may not be affected by forest edges in terms of their abundance and species richness because their species composition was found to be similar between forest edges and interiors in this study. Unlike many of the forest specialists, edge-associated species (e.g. Chlaenius naeviger) were found at 40 m into the forest interior in our study, indicating that these species penetrated deeply into the forest interior. This result was similar to that found by Molnár et al. (2001). Although further study into the edge effect in well-preserved forest areas may be needed, the effect of the forest edge on forest biota in young regenerating forests seems to be inappreciable.

Effect of landscape structure on ground beetles

In forest interiors, forest specialist ground beetles are influenced by various landscape factors such as patch sizes (Magura et al. 2010), and habitat stability with increasing forest age (Magura et al. 2015). In addition to these factors, forest connectivity, i.e. the connectedness between patches of suitable habitat, is also important (Fischer and Lindenmayer 2007). Although the forests studied in both regions were similar in terms of their forest type and age, the forests in Hoengseong were generally continuous and well preserved, whereas Hwaseong is a highly fragmented and developed landscape. This difference in forest connectivity may have resulted in the different patterns of species richness and species composition of ground beetles in the forest edges through elements such as varying levels of habitat heterogeneity, patchiness and stability. The size of the smallest patch in our study sites was 4.6 ha and, thus, our study patch sites were generally large, compared with other studies conducted in temperate forests (e.g. Fujita et al. 2008; Do and Joo 2013). For example, some large-bodied forest specialists (e.g. C. jankowskii jankowskii) were found from smaller fragmented patches (1.05–2.41 ha) in another study conducted in Korea (Do and Joo 2013). In our study landscape, however, smaller patches might not have contained an interior core habitat that could maintain forest specialists because most patches were not circular. According to Soga et al. (2013), core habitat may exist in small patches if the patch size is at least 1 ha and has a circular form. However, patches in actual landscapes have complex shapes, meaning there may be no core habitat in non-circular patches. We suspect that the absence of large-bodied and/or forest specialist populations in large patches in Hwaseong may indicate an absence of core habitats in that area.

In general, patterns in land-use at the landscape scale markedly affect the community structure of ground beetles producing distinct assemblages (Vanbergen et al. 2005). Decreases in forest area can negatively affect forest-associated species, whereas open-habitat species can expand their range of distribution. Many studies have investigated edge effects on ground beetle communities in natural habitats such as forest–grassland edges (e.g. Magura et al. 2001; Molnár et al. 2001; Magura 2002; Yu et al. 2007; Ewers and Didham 2008; Roume et al. 2011). These studies generally suggest that the forest edges and their surrounding areas operate as source habitats for the dispersal and re-colonization of forest-associated species after habitat disturbances; this is because forest specialists can freely move between the continuous forests and adjacent habitats. However, the pattern of species distribution across the forest–farmland edge in fragmented landscapes appears to be inconsistent with those of natural forest edges and clear-cut edges. This is probably the result of anthropogenic changes in soil properties (Koivula et al. 2004), habitat characteristics (Koivula et al. 2004; Vanbergen et al. 2005) and the presence of paved roads (Koivula and Vermeulen 2005; Do and Joo 2015) that can act as barriers to dispersal and colonization. Therefore, it is suspected that large or continuous forests in agro-forest landscapes may not function as source habitats for other populations in patches, especially flightless beetles. Nonetheless, Koivula et al. (2004) argued that open habitats and small forest fragments in an agro-forest landscape can accommodate several open-habitat species, such as endangered butterflies and dung beetles, and may therefore have potential uses in biodiversity conservation.

Changes in species distribution across the forest–farm edge

In this study, we examined the habitat affinities of several Korean ground beetle species, and found some interesting distribution patterns. For example, although C. naeviger has been reported as an open-habitat species in several previous studies (e.g. Ishitani et al. 2003; Jung et al. 2014), this species was caught mainly from the forest edge and neighboring sites in forest and non-forest areas. Thus, it is plausible that C. naeviger is actually an edge-associated species. P. jessoensis has also been cited as an open-habitat species according to some studies (e.g. El Sayed and Nakamura 2010; Jung et al. 2014), but this species was collected from the forest interior and edge of the PQF1 and PQF2 sites in Hoengseong. P. jessoensis is also known to be a hydrophilic beetle (Lake Biwa Museum 2014) and has generally been found in moist forests at other monitoring sites in Korea (unpublished data). Thus, P. jessoensis appears to be a generalist ranging from open habitat to moist forest interiors. Finally, the spatial distribution of C. smaragdinus branickii, known as a forest specialist and a very large-bodied species in Korea, is quite interesting. C. smaragdinus branickii was abundant at the forest interior and edge in Hoengseong, but was more abundantly collected from the forest edge and surrounding habitat in Hwaseong. In addition, adults of C. smaragdinus branickii in Hwaseong moved freely across the forest edge and neighboring habitats, but their larvae were generally restricted to the forest interior (data not shown). For these reasons, we suggest that C. smaragdinus branickii can actively move from forest to open-habitats to find prey, such as snails. This movement may assist farmers through the dispersal of predatory forest species into agricultural fields (Roume et al. 2011). This characteristic of C. smaragdinus branickii may also allow it to colonize in the forest patches of agro-forest landscape, unlike another similar large forest specialist, C. jankowskii jankowskii. Although more varied and detailed studies on the edge effect are still needed to evaluate ground beetles as well as other taxa, the preferred habitat types of each ground beetle species determined here will be useful for validating the results of future studies.

Conclusion

Ground beetle assemblages (species richness and composition) differed in the forest edges of two agro-forest landscapes, suggesting that ground beetles responded differently to the forest edge according to the landscape structure. Habitat fragmentation in agro-forest landscapes may restrict dispersal and re-colonization of forest specialists, and thus, forest specialist ground beetles may not disperse to adjacent agroecosystems and neighboring patches. This would decrease the biodiversity of ground beetles in agro-forest landscapes, especially in forest–farmland edges in highly fragmented landscapes. Therefore, to enhance biodiversity in forest–farmland edges, forest edges in the agro-forest landscape should be protected by expanding field margins through gradual edges and preserving core habitats in forest patches.