Zhenrong Yu, College of Agricultural Resources and Environmental Sciences, China Agricultural University, Beijing 100193, China. E-mail: firstname.lastname@example.org
Abstract. 1. In the intensively cultivated major crop production areas of China, large knowledge gaps still prevail with regard to the current status of biodiversity in general, and especially in relation to agriculture management and planting patterns. Effective measures for species conservation are hence widely lacking.
2. The diversity of ground beetle (Coleoptera: Carabidae) assemblages was compared between intensively managed wheat/maize fields, cotton monocultures, lawns, orchards and semi-natural woodland located in Quzhou county, a typical, intensively managed agricultural region in the North China Plain.
3. Although significant differences were found in species composition between different habitats, diversity of ground beetles in non-cropping sites was not significantly higher than in intensively managed wheat/maize double-cropping fields, while cotton monocultures had a significant lower carabid diversity. A combination of intensively managed wheat/maize double-cropping fields with orchards appears to preserve substantial proportions of ground beetle diversity.
4. To conserve the full spectrum of beetle species currently observed in the agricultural landscape of the North China Plain, the creation of both diverse habitats and a diversity of cropping systems are important, which includes a strengthening of less intensive farming practices.
While intensification of agricultural activities has been identified as a major driver of recent arthropod species loss (Matson et al., 1997; Tilman et al., 2001), some studies also indicate that several species have become adapted to agricultural management (Callaham et al., 2006; Saska et al., 2007). In the intensively cultivated major crop production areas of China, large knowledge gaps still prevail with regard to the current status of species richness in general, and especially in relation to agriculture management and planting patterns. Measures for species conservation in this intensively cultivated landscape are hence widely lacking. Only the investigations into carabid biodiversity in the desalinised agricultural landscape of Quzhou by Liu et al. (2006) and the study by Zhiping et al. (2006) into the impact of agricultural management on earthworm diversity in Huantai allowed limited preliminary insights.
In this study, we set out to investigate the prevailing knowledge gaps, investigating the general status of diversity and the role of the small remnants of non-cropped area in sustaining and enhancing species density using carabid beetles as indicator taxon. Carabid beetles were selected as they are relatively well-known both taxonomically and ecologically (Lövei & Sunderland, 1996; Niemelä, 1996). They also have been successfully used as bioindicators in a variety of studies (e.g. Thiele, 1977; Humphrey et al., 1999; Magura et al., 2000; Vanbergen et al., 2005; Liu et al., 2006), as they are susceptible to standardised sampling via pitfall trapping, easy to preserve, and they react sensitively to changes of their environment. With most species being predominantly insectivorous, they also play an important potential role in pest control in the agrarian landscape (Kromp, 1999). We hypothesis that: (1) the small remnants of non-cropped area sustain greater diversity than cropping areas, thus enhancing the overall diversity in the intensively cultivated agro-landscape. (2) Habitats not used for agriculture harbour a significantly different species assemblage of carabid beetles, hence overall enhancing the local species pool.
Materials and methods
The study was conducted at the Agricultural Research Station of the China Agricultural University at Quzhou County (36°20′N, 114°00′E). This county represents a typical agricultural area in the North China Plain, characterised by intensive agricultural production with a mosaic of small fields dominating the landscape, interrupted only rarely by semi-natural habitats like hedges and woodland patches. In 2000, the county was principally covered by agricultural land (76.2% of the land area), followed by residential area (12.5%), water (6.0%), infrastructure (2.2%) and waste land (1.7%), with only very small proportions remaining for orchard (1.0%) and woodland (0.4%). Within the agricultural land, wheat/maize rotation accounted for 43% of the area and cotton monoculture for 20%. Rotation of winter wheat and a range of summer crops amounted to another 30%, and other crops such as oil seeds, vegetables or melon covered 7% (Handan Statistical Bureau, 2001).
Apart from a mosaic of agricultural fields, the Agricultural Research Station also comprises a small planted woodland area which was established more than twenty years ago, and various patches of lawn created for their amenity value. The woodland area is a representative of mainly native woodland flora in the county where this habitat is otherwise virtually non-existent because of the transformation of land for agricultural use. At all three woodland plots Populus tomentosa was the dominant canopy tree. Although the selected woodland is not being used for timber production, some management was applied to this habitat. In 2005, irrigation measures were taken to ensure a better growth of the tree specimen and weeds were manually removed in the undergrowth to prevent the manifestation of large populations of insect pest species in the woodland. The lawns contained mainly native grassland species (including herbaceous species Trifolium repens and Euphorbia humifusa) and were also regularly cut and irrigated during dry periods. The vegetation at the lawns was cut at regular intervals and hence experienced considerable physical disturbance. Nonetheless, both woodland and lawn habitats experienced much lower overall levels of anthropogenic disturbance than the agricultural areas in their vicinity, especially in relation to the application of chemicals, but also to ploughing and soil compaction by the use of heavy machinery.
Both woodland (W) and lawns (L) were selected for carabid sampling, in addition to three different habitats representing common planting systems intensively managed for maximum production: wheat/maize double-cropping fields (WM), cotton monocultures (C) and orchards (O).
In the agricultural areas, the total amount of nitrogen applied in each growing season was 300 kg N ha−1 at the wheat/maize rotation fields, 100 kg N ha−1 at the cotton fields and 110 kg N ha−1 for the orchards, with additional herbicides applied 1–2 times in the wheat/maize rotation fields. Pesticide applications during each growing season reached 8–10 times in the orchards, 1–2 times per crop in the wheat/maize double-cropping fields and high frequencies in cotton fields, with five applications in 2005 and weekly applications starting from the early growth stages of the plants in 2006.
Four spatially distinct replicate plots were selected for each agricultural habitat, and three spatially distinct replicate plots in the woodland and lawns (Fig. 1). Plots measured 20 × 20 m2, each. Pitfall traps were used to sample carabid beetles as a very convenient and easy-to-operate method (Greenslade & Greenslade, 1971) yielding highly standardised samples (Thiele, 1977; Southwood, 1978). When interpreting the results, it has nonetheless to be kept in mind that pitfall trap samples reflect species’ activity patterns rather than their actual abundances (see also Briggs, 1961; Mitchell,1963; Greenslade, 1964; Luff, 1975; Baars, 1979; Pekár, 2002). A total of eight pitfalls were used for carabid sampling on each plot. These traps were placed at a distance of 4 and 7 m from the plot centre along two N-S and E-W facing diagonal lines intersecting in the middle of the plot. The pitfall traps consisted of plastic beakers of 8 cm diameter and 11.5 cm height, which were partly filled with 75% alcohol to kill and preserve the specimens (Southwood, 1978). Each trap was protected from rain by a simple aluminium roof positioned ca. 5 cm above the trap. Sampling took place from the beginning of May to early October in both 2005 and 2006 over a total number of 120 sampling days. To better preserve the beetles, pitfall traps were emptied at the fifth day of sampling in each month and the sampled specimens transferred into fresh alcohol before the traps were re-set filled with new alcohol to be emptied again on the tenth day.
To evaluate the effects of planting systems and related management practices on the ground beetle diversity, a set of environmental parameters including plant species richness and soil properties (upper 10 cm of the mineral soil) were investigated (Table 1). Plant species were recorded once in June and once in September on each 20 × 20 m2 plot in 2005, and the number of plant species (NPS) was counted. The soil properties analysed in this study included pH values and soil organic matter (SOM) contents, plant-available phosphorus (P) and total nitrogen (N). As the study area was formerly affected by high levels of salinity (Xin & Li, 1990), the soil salt content (measured as conductivity – EC) was also recorded.
Table 1. Environmental parameters recorded at the different habitates (mean ± SD).
Number of Plant species
Soil organic matter content (%)
Plant-available phosphorus and total nitrogen (ppm)
Soil salt content (uS cm−1)
14.6 ± 2.6
1.58 ± 0.30
0.330 ± 0.076
0.089 ± 0.011
219.5 ± 48.86
8.0 ± 0.1
16.5 ± 4.1
1.77 ± 0.09
0.412 ± 0.202
0.085 ± 0.020
210.1 ± 17.39
8.1 ± 0.1
32.5 ± 4.0
1.71 ± 0.09
0.268 ± 0.083
0.074 ± 0.018
162.3 ± 32.34
8.1 ± 0.1
19.7 ± 4.0
1.86 ± 0.35
0.423 ± 0.138
0.087 ± 0.013
201.8 ± 74.10
7.9 ± 0.2
34.7 ± 2.1
1.89 ± 0.36
0.187 ± 0.012
0.096 ± 0.020
126.7 ± 12.07
8.1 ± 0.0
Specimens caught at the same plot during sampling season were pooled for statistical analysis. The Berger-Parker Dominance Index was calculated for all ground beetle samples to analyse the degree of dominance of the most common species. This index represents the proportion of the most dominant species in the overall sample, with high levels of dominance of a single species generally representing low levels of diversity (Caruso et al., 2007). To explore differences in the alpha-diversity of carabid communities between treatments further, Fisher’s α and Hurlbert rarefaction (Hurlbert, 1971) were used. Both these methods are known to be applicable for samples of strongly varying sample sizes (Axmacher et al., 2004a,b). Rarefied species numbers and 95% confidence intervals of the respective values were calculated using Analytic Rarefaction 1.3 (Holland, 2008) for the largest common sample size of all habitats (83 individuals; all samples of the same habitat type were pooled due to the low number of specimen collected especially at woodland and orchard plots). Fisher’s α was calculated using EstimateS 6.0b1 (Colwell, 2000).
A one-way anova and post hoc tests (Duncan’s multiple range test) were carried out for Fisher’s α and the Berger-Parker-Index to evaluate if significant differences exist between the diversity of carabid species assemblages recorded at different habitats. This analysis was conducted with the program Data Processing System (DPS) V3.11 (Tang & Feng, 1997, 2002).
Non-linear multidimensional scaling (NMDS) of the chord-normalised expected species shared (CNESS)-index of dissimilarity (Trueblood et al., 1994) was used to analyse species turnover rates of carabid ensembles between sites. CNESS is a metric version of Grassle and Smith’s NESS similarity index (Grassle & Smith, 1976). It is one of the most appropriate indices for analysing quantitative data (Trueblood et al., 1994), allowing the calculation of a probability-based measure of similarity between samples which can be of differing sample sizes. The sample size (parameter m) for which the similarity matrix is calculated can be varied, but must be kept at or below the maximum sample size of the smallest sample. The CNESS dissimilarity matrix was derived using the program COMPAH (Gallagher, 1998), while the NMDS was calculated using Statistica (StatSoft Inc., 2001). Finally, canonical correspondence analysis (CCA) was employed using PC-Ord (McCune & Mefford, 2006) to explore relationships between environmental parameters and the composition of carabid assemblages.
Species composition and alpha-diversity
A total of 1055 carabid individuals representing 28 species were caught in the agricultural landscape over a period of 2 years (Table 2). The most common species were Chlaenius micans (Fabricius) and Scarites terricola Bonelli, accounting for 23.1% and 18.7%, respectively of all individuals caught. Different species reached dominance at different habitats. C. micans (54.5%) dominated the carabid assemblages at cotton fields, S. terricola (40.3%) at the lawns, Harpalus bungii Chaudoir (26.3%) and Asaphidion semilucidum (Motschulsky) (12.5%) in orchards, S. terricola (27.7%) again in wheat/maize rotation fields, and H. bungii (33.3%) in the woodland habitat.
Table 2. Overview of individuals trapped at different habitats in the agricultural landscape of Quzhou county in 2005 and 2006.
The Berger-Parker Dominance Index (BPI) was greatest at cotton fields (mean = 0.52, SD = 0.16) and lowest at the orchards (mean = 0.27, SD = 0.12) (Table 3), indicating greater dominance in single species at cotton fields in comparison to the other habitats. Accordingly, Fisher’s α reached highest values at the orchards (mean = 6.47, SD = 1.8) and lowest values at the cotton fields (mean = 3.15, SD = 1.33) (Table 3), indicating greatest diversity in orchards and lowest in cotton fields, with the remaining habitats characterised by intermediate values for both indices. Woodland diversity (BPI: mean = 0.33, SD = 0.08; Fisher’s α: mean = 5.79, SD = 3.08) in particular reached values very similar to the ones observed in orchard samples, but showing a much greater variety especially in relation to Fisher’s α. Intensively used wheat/maize double-cropping (BPI: mean = 0.36, SD = 0.09; Fisher’s α: mean = 4.79, SD = 0.80) yielded communities of only slightly lower diversity, with diversity levels at the lawns (BPI: mean = 0.43, SD = 0.12; Fisher’s α: mean = 3.78, SD = 1.24) much more resembling the ones observed in the cotton monocultures. Statistically significant differences were nonetheless only found between orchards and cotton fields for both Berger-Parker Dominance Index and Fisher’s α (BPI P = 0.024; Fisher’s αP = 0.038). Similar to Fisher’s α, rarefied species numbers for 83 individuals (Table 3) reached the greatest values at orchards and at wheat/maize double-cropping fields (mean = 15.1), and the lowest values at cotton fields (mean = 10.6). Again, the woodland samples (mean = 14.9, SD = 0.26) reached similar rarefied species numbers to the orchard and wheat/maize plots, with lawn samples (mean = 12.0, SD = 1.57) resembling the low diversity at cotton monocultures (mean = 10.6, SD = 1.53).
Table 3. Berger-Parker dominance index and alpha-diversity measured as Fisher’s α and rarefied species number at the different habitats in combined samples from 2005 and 2006 (for BP index and Fisher’s α, values show mean ± SD, with different letters indicating significant differences according to Duncan’s multiple range test (95% confidence interval) BPI: d.f. = 4, F = 2.56; Fisher’s α: d.f. = 4, F = 2.33, for rarefaction, values indicate mean and 95% confidence interval, with different letters again indicating statistically significant differences).
Rarefaction (n = 83)
Different letters indicate significant differences according to Duncan's multiple range test (95% confidence interval).
0.52 ± 0.16a
3.15 ± 1.33b
0.36 ± 0.09ab
4.79 ± 0.80ab
0.27 ± 0.12b
6.47 ± 1.80a
0.43 ± 0.12ab
3.78 ± 1.24ab
0.33 ± 0.08ab
5.79 ± 3.08ab
Two-dimensional scaling of the CNESS index of dissimilarity for a minimum sample size parameter m = 1, emphasising similarity in dominant species (Fig. 2a), indicated that assemblages originating from the same habitats formed distinct clusters separately from assemblages of other habitats. The only exceptions to this observation were orchard and woodland plots, which formed a large, joined cluster. Overall, the ordination plot indicated a pronounced shift in the dominant carabid species between different habitats, with assemblages at wheat/maize fields taking an intermediate position between cotton field- and lawn-assemblages on one, and woodland- and orchard-assemblages on the other side. The latter also showed large plot-to-plot variations in the composition of dominant species.
The NMDS ordination plot for an intermediate sample size parameter (m = 10) (Fig. 2b) showed very similar patterns already observed for m = 1, with the exception that in this case all carabid assemblages of the same habitats formed distinct, separate clusters, which meant that in relation to less common species, the composition at woodland sites showed a distinguishing level of dissimilarity to orchard sites, which took an intermediate position in the ordination plot between woodland plots and lawns as well as wheat/maize plots.
In the analysis of relationships between ground beetle community structure and environment factors comprising of the six soil properties and plant species richness, the first two axes of the CCA ordination plot explained 26% and 12% of the total variance, respectively. Monte-Carlo permutation tests furthermore indicated that environmental variables had a significant effect on changes in the beetle community structure (998 runs, P (1st axis) = 0.011).
Figure 3 represents the biplot of environmental variables and carabid species scores. The first axis is strongly related to the number of plant species (biplot score: 0.58), as well as to soil salt content (biplot score: −0.41). The species H. bungii (S5), A. semilucidum (S9), Pterostichus gebleri Dejean (S10), Amara sp. (S20), Stenolophus sp. (S23), Peronomerus auripilis Bates (S26), Harpalus tinctulus (Bates) (S27) and Cymindis daimio Bates (S28) all had large positive species scores along this axis, indicating an occurrence of these species at plots with a high diversity in plant communities and low soil salt contents. Conversely, C. micans (S1) had a large negative species score along the first canonical axis, thus occurring at plots with high soil salt contents and a lower phytodiversity.
The second axis is strongly related to the soil contents in plant-available phosphorus (biplot score: 0.32), salt contents (biplot score: 0.24) and pH (biplot score: −0.19). P. gebleri (S10), Amara sp. (S20), Stenolophus sp. (S23), H. tinctulus (S27) and C. daimio (S28) had large negative species scores along the second canonical axis, with their occurrence coinciding with sites characterised by comparatively low amounts of plant-available phosphorus and salt in the soil, as well as by high pH values. S. terricola (S2), Harpalus simplicidens Schauberger (S12), Pterostichus microcephalus (S21) and Panagaeus davidi Fairmaire (S24), finally, were positioned far on the positive side of the second axis, hence occurring at sites characterised by high soil contents in plant-available phosphorus, salt and low soil pH values.
An increase in habitat heterogeneity has been reported to be an efficient way to enhance biodiversity, for example by preserving or establishing woodland habitats (Moore et al., 2003; Fraser et al., 2007), hedgerows (Pollard & Holland, 2006), extensively used field margins (Woodcock et al., 2005; Yu et al., 2006) or areas of fallow land (Buskirk & Willi, 2004; Macdonald et al., 2007). In agreement with this assumption, field margins in Quzhou County have also been shown to harbour a greater diversity of ground beetles than neighbouring farmland (Liu et al., 2006). In contrast to these findings, the semi-natural woodland habitat investigated in this study harbours an astonishingly low alpha-diversity, with alpha-diversity of the respective species assemblages not exceeding those recorded at orchards nor even reaching significantly higher levels than at very intensively managed wheat/maize rotation fields. Even in comparison to intensively cultivated cotton monocultures with a very high frequency of pesticide applications, both woodland and especially extensively managed lawns contained only slightly higher levels of carabid diversity.
However, when assessing larger-scale biodiversity conservation, beta diversity, expressed as species turnover between habitats, indicated a distinct difference of assemblages amongst woodland sites and all other habitats. Moreover, samples from woodland and lawn sites added an additional three species, Harpalus crates, Harpalus sinicus and C. daimio to the species list, which represents about 10% of the overall species pool. These species were very rare, so that they could have simply been missed during sampling at the other plots despite the high intensity of sampling. It is hence not completely certain if the species are generally rare, or if they are indeed ‘true’ non-cropland species. It can be stated that overall, habitats in the study area which have not been used for agricultural purposes did not contain a higher diversity of carabid species than some of the very intensively managed habitats, but the further degradation or destruction of these habitats will most probably result in the loss of their unique proportion of the ground beetle species pool, leading to a lowering of gamma-diversity in the agricultural landscape.
One reason for the low biodiversity encountered in the lawn habitat despite the low level of anthropogenic influence for example in relation to chemicals applied could be the uniformity in plant species composition at these sites (Table 1), since the lawns originated from a seed mixture of durable grass species which hence accounted for the vast majority of the vegetation cover at these sites. Furthermore, the lawn was by far the most homogeneous habitat in relation to vegetation height and structure. In the woodland, the removal of weeds could potentially have a negative effect on the species richness of carabid assemblages by again increasing the habitat homogeneity. The relatively high biodiversity encountered at the orchard despite the intensive use of pesticides at these plots may be related to the combination of a high plant species richness (Table 1) and a complex vegetation structure.
Another reason for the surprisingly similar levels of biodiversity in the extensively managed, semi-natural habitats and that of wheat-maize fields may be that only very small remnants of semi-natural habitats exist in the study area, so that populations of, and colonisation rates by woodland-specialists will be very low. Size of habitat patches, plant species composition (Thomas & Marshall, 1999; Meek et al., 2002), habitat connectivity (de la Peña et al., 2003; Burel et al., 2004) and surrounding habitat type (Öberg et al., 2007) are all well-known factors affecting species composition and diversity of carabid assemblages. Very small remnants of semi-natural habitats scattered in a highly crop-dominated landscape will reflect on all of these factors. Lawn plots, ranging from about 500 to 800 m2 in area, and woodland sites ranging from about 2600 m2 to 4500 km2 are both effectively isolated from similar habitats by wide tracks of fields as well as by concrete pathways and roads, making it virtually impossible especially for flightless species to reach these habitat islands.
On the other hand, the relatively higher diversity in the cropping sites was consistent with the argument that biodiversity conservation and agricultural landscapes do not need to be incompatible and agricultural habitats if appropriately managed may even supports a substantially higher biodiversity than pristine habitats (Pimentel et al., 1992; Tscharntke et al., 2005; Ameixa & Kindlmann, 2008). Planting systems and management practices may provide a further explanation for the observed distribution pattern of carabid assamblages in this study. The canonical correspondence analysis indicated that the number of plant species, the content of plant-available phosphorus as well as soil pH values were all significantly correlated with the distribution patterns of carabid species. The effects of plant species diversity has already been mentioned above, and the later two factors may indirectly affect beetles by impacting the growth of plant, thus modifying the microhabitat (Cárcamo & Parkinson, 1999). All three parameters are greatly shaped by farming practices linked to the respective planting systems, for example by changes in fertiliser levels applied, and by management directly impacting the diversity of the plant communities, for example mowing or weeding. Moreover, specific management practices related to the planting systems can serve to explain the relatively high diversity of carabid on farmland as compared to more ‘natural’ habitats. For example, pesticides in the orchards being applied directly onto the fruit trees may have allowed the persistence of shielded safe harbours for ground-living arthropods, which potentially allowed them to remain disaffected by the chemicals. The high biodiversity recorded in the very homogeneous winter wheat/summer maize rotation fields might also relate to less intensive pesticide applications and greater vegetation covers early in the year (Hance, 1990; Booij & Den Nijs, 1992; Booij & Noorlander, 1992). Another potential reason might be the long history and dominance of cultivation of both wheat and maize in the study area that may have resulted in the current species pool being composed of species partly adapted to this planting system, a pattern also recorded in the agricultural landscape elsewhere (e.g. Callaham et al., 2006; Saska et al., 2007).
In relation to conservation measures in the agricultural landscape, this research indicates that habitat diversity appeared to strongly influence biodiversity levels in this intensively managed agricultural landscape: the extensively managed orchards as well as cropping fields can harbour diverse assemblages of ground beetles, but the semi-natural woodland and extensively used lawn widely relates to the small additional component of the local beetle fauna that was exclusively recorded at these two habitats. As plant diversity was a strongly influential factor in species turnover and distribution, future management especially of less intensive land-use forms like orchards should hence be aimed towards preserving a greater plant diversity to help improving also arthropod species richness in intensively managed farming areas. Farm management which directly impacts plant species assemblages like pesticide applications, weeding and mowing, but also factors influencing them more subtle like fertilisation, all appear to impact the distribution and diversity of carabid beetles. The management and structure in the non-agricultural habitats in particular should be carefully aligned to ensure an optimum efficiency of biodiversity conservation in the very intensively managed landscape.
When interpreting the results of this study, it is finally important to keep in mind that the overall diversity of ground beetles (1055 individuals and 28 species) was poor, considering that 144 traps were used for a total of 120 sampling days over two 6 month-periods in two consecutive years distributed over five distinctly different habitat types. This is in strong contrast to sampling success of other studies, with e.g. Saska et al. (2007) recording 75 species and 11 000 specimens when sampling with 168 traps over 21 sampling days at six different habitats in the agricultural landscape and Menalled et al. (2007) recording 1609 individuals comprising 33 species when using 120 traps at four different agricultural habitats over 42 sampling days. Previous research at Quzhou county in both 1996 and 1997 using 150 traps for 48 sampling days at 30 different plots and another 100 traps for 30 days on 20 plots in 1997 also emphasised the very low carabid diversity levels at Quzhou, with only 670 individuals representing 19 species being recorded (Liu et al., 2006). Further research is needed into the general causes of these low diversity values, with the historically very high soil salinity (Den Boer, 1970) and other biogeographical constraints like the overall extremely homogeneous land-use patterns and long history of homogeneous agricultural use in the North China Plain being potential candidates.
An increase in habitat heterogeneity by establishing or increasing the proportion of non-cropping vegetation cover is widely seen as an efficient measure to enhance the biodiversity in the agrarian landscape. In Quzhou County, carabid assemblages showed different compositions amongst different cropping and non-cropping sites, indicating that both diverse habitats and a diversity of cropping systems are important in sustaining high biodiversity levels in this intensive agricultural landscape; with non-cropping areas adding additional components of the local beetle fauna. Meanwhile, agricultural land, especially with slightly ameliorated management practices, can play an important role in sustaining species diversity levels in an intensively cultivated agricultural landscape.
The authors are greatly indebted to the National Natural Science Foundation of China (30570318 and 30800150) and the National Key Technologies R&D Program of China (2006BAJ10B05). We are also very grateful to Dr Liang Hongbin from the Institute of Zoology at the China Academy of Sciences for his great help with the identification of carabid specimens.