The research was done at three sites in the Czech Republic from May to July 2013. The first site was an alfalfa field (49°34.41'N, 17°17.17'E) on the border of the town of Olomouc. This large field of ca 250 square metres is surrounded by other fields (with cereals) and a railway embankment. In the previous year it had also grown alfalfa. The field is under conventional management including use of herbicides and ploughing.

The second site was an old meadow (50°26.85'N, 15°0.00'E) being mown once to twice per year for the last 30 years. This meadow of ca 500 square metres is surrounded by fields and gardens with mixed wood across the road and is ca 6 km north-east of the town of Mladá Boleslav. The third site studied was a mixed forest (49°15.66'N, 17°17.72'E) 6 km south-west of the town of Kroměříž. The forest is classified as Fageto-Quercetum illimerosum trophicum; dominant trees are oaks, hornbeams and some pines, with Rubus fruticosus, Galium odoratum and Galium aparine as dominants of undergrowth. The soil surface of this forest is covered by a rather thick layer of oak leaf litter.

In the Czech Republic generally, the weather conditions during the study period were characterised by average or slightly increased temperatures and higher than average precipitation in May-June, and a very hot July in contrast to long-term average values. The previous winter season was rather warmer and with higher precipitation (ref. historical territorial data at www.chmi.cz).

Soil macrofauna, including millipedes and centipedes, was sampled using three methods at each site. Pitfall traps (10 traps consisting of glass jars with inserted plastic pots of diameter 7.5 cm filled with 2 dl of 4% formaldehyde in water with some detergent, metal covers) were arranged in 2 lines of 5 traps with a span of 10 m, and inspected at 2-week intervals. Five soil samples (25 × 25 × 10 cm including litter layer) were obtained using a spatula, three times per study (i.e. 15 soil samples per site) and transported to the laboratory in plastic bags. Hay-bait traps were made from a wire gauze (2 cm mesh) shaped as a simple pocket of size 25 × 25 cm. Each pocket was marked by a code written on the band. These pockets were filled with hay (commercial hay mixture for feeding rodent pets) and submerged into water for 2 hours before installation. Altogether, 60 hay-bait traps were placed horizontally at each site in a following scheme: 5 lines of 12 traps (2-5 cm under soil surface) over a length of 2 m with 10 m between lines. All traps were installed at the same time and 5 traps were taken away each week during the course of the study lasting for 12 weeks. Hay-traps were transported into the laboratory inside separate plastic bags.

Soil samples and hay-traps were heat-extracted immediately in the laboratory using simple Kempson devices (Tuf and Tvardík 2005). Hay-traps were extracted for a week, soil samples for 2 weeks, both under electric 60W-bulbs. Extracted animals from both soil samples and hay-traps were sorted to higher taxonomic groups and millipedes and centipedes were identified to the species level.

We tested the effects of trapping time and methods on species richness by repeated-measures on traps with nested design. The traps were nested in each of the three study sites (field, meadow, forest). Explanatory variables in the model were trapping time and trapping methods. The response variable was defined as a number of species per trap for particular time and place. Habitat type was used as random variable. We used a mixed model to estimate the correct error term and degrees of freedom. To test this effect, a generalized linear mixed model (glmmPQL, part of R package MASS) was used with negative binomial error distribution and log link function (Bates et al. 2014).

To test if one level of a particular factor (trapping method and study site) is more variable than other levels of the same factor, a permutation test was used (permutest.betadisper, part of R package vegan). This permutation based method tests pairwise comparisons of group mean dispersions. It is based on the t-statistic computed on pairwise group dispersions. A distance matrix was computed based on “Bray-Curtis” index of dissimilarity (vegdist, part of R package vegan). Then the function “betadisper” (part of vegan package in R) was used to calculate variance for each group of samples. Variance was computed as average distance of group members to the group centroid.

Rarefaction curves were constructed to show how the species richness varies for the same sample size between the three trapping methods. Function “rarefy” (part of vegan package in R software) computed the expected species richness and standard deviation in random subsamples of a particular sample size from the community. Data were analysed using R software (R Development Core Team 2011).

If we are interested in using hay-bait traps as an adequate (or even better) substitute for soil sampling, we have to compare species lists of millipedes and centipedes trapped by these methods. There were only three species recorded exclusively from soil sampling, i.e. missing in hay-bait traps: millipede Brachydesmus superus and centipedes Geophilus electricus and G. truncorum. The minute millipede species lives preferably in clay soils with litter (Lee 2006) usually in huge quantities. This species is a dominant species recorded by pitfall traps in cities (Riedel et al. 2009), so its absence in pitfall traps at the meadow site is probably caused by its low abundance. Geophilus truncorum was recorded once only, so it is hard to evaluate effectiveness of sampling of so “rare” a species. Nevertheless, both geophilomorphs (G. electricus and G. truncorum) are known as predators of earthworms (Sergeeva et al. 1985, Keay 1986); for this reason, hay-bait may be not attractive for them as they follow earthworms into their corridors in soil. So, to collect G. electricus, soil sampling or direct hand collecting seems to be necessary. Other geophilomorphs (C. flavidus, S. nemorensis, S. acuminata, G. flavus) are common species, which are frequently found by individual hand-collecting; they live in soil near the surface, under logs, bark and stones (e.g. Barber and Keay 1988). Their presence in shallow hay-bait traps is not surprising as these species were sampled by pitfall trapping too.

One millipede species, Julus scandinavius was recorded exclusively in a hay-bait trap, but as one specimen only was found no generalization can be made. Many more species were found in both hay-bait traps and pitfall traps but not in soil samples. Nevertheless, hay-bait traps are not a substitute method to pitfall trapping as there were significant differences between species lists recorded by these methods (see Tab. 2), but it can definitely substitute the soil samples.

Centipedes, and especially millipedes, are attracted into the hay-bait traps. The possible reason can be as a food source and/or sustainable shelter with higher humidity. At least for millipedes, food source seems to be the more probable explanation; wet cloth method (offering higher humidity) did not record any millipedes in African savannah ecosystems (Druce et al. 2004). More probably, millipedes and centipedes are attracted by food availability, as it can be associated with hay decomposition and colonization of the traps. Smaller decomposers colonising baits are welcomed food for carnivorous centipedes (e.g. Perry et al. 1997).

Eight to ten weeks seems to be the optimal exposure time for hay-bait traps in Central European conditions. A similar result was found by Ožanová (2001) using grass traps (a small heap of mowed grass on the surface of meadow), with a much higher number of species after 7 weeks than for a shorter exposure time. Although Prasifka et al. (2007) did not evaluate the effect of length of exposure time of bait traps, it is evident from their results that below-ground traps were more effective after 8 weeks than after 6 weeks. It supports our results that the best length of exposure of bait traps is from 8 to 10 weeks, although we aware of difficulties with this generalization. The best length of exposure is not dependent only on a type of habitat, but also on climate conditions (rainy or dry weather) and time of year when exposed. Traps installed in Central European conditions in late autumn or in winter or during dry hot summer can be colonised in different ways as invertebrates change their activity and position in the soil profile during the year (David 1984, Geoffroy 1985, David et al. 1996, Tuf 2002). The general recommendation for using these traps when installed in spring is to use them for 8–10 weeks. Timing of installation and the length of time exposed in field sites will need to coordinate with agricultural activities such as sowing and harvest times.