S. Mahunka szerk.: Folia Entomologica Hungarica 56. (Budapest, 1995)

RESULTS AND DISCUSSION The hand-hold suction sampler proved to be efficient and useful to obtain quantitative samples of the arthropod fauna in different habitats. The air speed of the sampler over the 0.01 m

2 sampling area was 16.7 m/s. This compared very favourably with the 3.5 m/s speed of air flow of D-vac over a 0.1 m

2 sampling area (both figures are calculated from specifications provided by the manufacturers). In the parallel samples made with the suction sampler, pitfall traps and sweep net a total of 3298 animals were caught, which represented 24 higher taxa (Table 1). A twoway multivariate analysis of variance (Füstös et al. 1986) carried out on the number of animals caught in the various taxa by habitat and method showed the significance of both factors, with method exerting a greater effect on catches (Table 2). The interaction between method and habitat was also significant suggesting that the various sampling techniques had different taxon specific efficiency in the structurally different two habitats. In the suction samples 18 out of the total 24 higher taxa were represented. A principal component analysis carried out on samples by taxa (standardised catches) provided evidence that all three methods accessed a somewhat different set of the animal community present in the habitats sampled (Fig. 2). Both suction sampling and sweep net collected animals from the herbaceous vegetation, but suction sampling added animals of the lower vegetational strata and the ground surface, as well (e.g. Coleoptera, especially ground beetles, many spider families). Sweep netting was probably more efficient for collecting animals capable of fast movement and living in the higher vegetational strata (e.g. Orthoptera), whereas suction sampling seemed to be more efficient for light flying insects (e.g. Diptera and Hymenoptera). While there was much overlap between catches by suction and sweep net sampling, the PCA showed that pitfall trap catches were markedly different in the proportions of the different taxa caught. Although suction sampling collected animals moving on the ground surface, night active animals and animals that normally live hidden in the highest strata of the soil, in cracks, under stones, etc. were inaccessible to this method. However, these groups were well represented in pitfall catches (e.g. Coleoptera larvae, Collembola, Halticinae, Opilionidea, Isopoda, Chilopoda, Blattidae). This great difference in the spectra of target organisms makes the use of pitfalls inevitable despite many recent criticism which was aimed at the difficult interpretability of Table 2. Result of a two-way multivariate analysis of variance performed on the number of animals caught from the different taxa by habitat and by collecting method. The analysis was made on catches logarithmically transformed, and only 8 major taxonomic groups are included: Orthoptera, Heteroptera, Auchenorrhyncha, Aphidoidea, Diptera, Coleoptera, Hymenoptera and Araneae, for which ANOVA assumptions hold (The same test performed on untransformed data for nearly all taxa - which had an overall mean density 1 - gave the same qualitative results) Effect Wilks' Lambda Rao's R d.f. 1 d.f. 2 P Habitat 0.12 11.97 8 13 0.0001 Method 0.012 13.00 16 26 0.00001 Interaction 0.16 2.46 16 26 0.0