Ef fi ciency of pitfall traps with funnels and / or roofs in capturing ground-dwelling arthropods

Pitfall traps are widely used for sampling ground-dwelling arthropods. Their sampling effi ciency is affected by several factors, e.g. material, size and modifi cation of parts of the trap and sampling design. Pitfall trap sampling is also affected by the accumulation of plant litter in the traps, rain fall and by-catches of small vertebrates, which may cause a bias in the catch by obstructing traps or attracting certain insects. A roof that prevents rain and plant litter entering a trap, prevents dilution of the preservative and escape of arthropods. The main goal of present study was to compare the effect of four types of differently combined funnel and roof pitfall traps on the capture effi ciency of epigeal arthropods. We found that a funnel and/or a roof had no effect on spider catches. Total abundance of large carabids and thus the total abundance of ground beetles was lower in funnel pitfall traps without a roof than in other types of traps. However, funnel pitfall traps with roofs collected signifi cantly more carabid beetles, especially individuals of those species that are large or good fl iers. We conclude that funnel pitfall traps with roofs have no negative effects on capture effi ciency of ground beetles and spiders, therefore application of this sampling technique is strongly recommended.

which can be adjusted.Other factors are specifi c to the environment like habitat structure (Melbourne, 1999;Lang, 2000) or temperature (Saska et al., 2013 ) and the target group of invertebrates (like their typical abundance, activity and catchability) (Southwood & Henderson, 2000) and cannot be changed by the investigator.The optimal trap effi ciently collects invertebrates with minimal bias and reduce by-catches of non-target animals (New, 1999;Buchholz et al., 2011;Lange et al., 2011).The aim of the present study was to compare the sampling effi ciency of different types of pitfall traps.We specifi cally tested how the addition of a funnel and/or a roof to the traps affects the catches of two invertebrate predator taxa, carabids and spiders.We hypothesized that the use of funnel pitfall trap with a roof does not have a negative effect on the effi ciency with which it captures invertebrates.

MATERIAL AND METHODS
This study was conducted in 2014, from May to October in a forest-wet meadow complex at Turjánvidék, Central Hungary (46°43´N, 19°20´E).The climate is continental; the annual precipitation is a 500-600 mm and the mean annual temperature is 10-11°C (Bíró et al., 2013;Tölgyesi et al., 2015).Grasslands were mowed once a year, at the beginning of July.The dominant species of trees in the forest patches include narrow-leafed ash (Fraxinus angustifolia) and English oak (Quercus robur).
We compared four types of pitfall traps: (1) conventional pitfall trap without a roof, (2) conventional pitfall trap with a roof,

INTRODUCTION
Choosing the most effi cient method of sampling is crucial in studies aiming to compare invertebrate assemblages (Ernst et al., 2015).Pitfall traps are the most widely used tools for sampling ground-dwelling arthropods in ecological studies and monitoring programs (Southwood & Henderson, 2000;Babin-Fenske & Anand, 2010;Da Silva et al., 2011;Isaia et al., 2015;Brown & Matthews, 2016).Pitfall traps are easy to transport and install; they cost little, cause relatively low disturbance and can yield a large number of individual invertebrates and species (New, 1998;Woodcock, 2005;Santos et al., 2007).Usually, traps consist of plastic or glass containers that are dug into the ground with the top fl ush with ground level (Brown & Matthews, 2016).In most studies, traps contain a preservative fl uid to prevent arthropods from escaping and preserve the material collected (Jud & Schmidt-Entling, 2008;Knapp & Ruzicka, 2012).However, there are various types of pitfall traps, which also incorporate barriers, drift-fences, funnels, roofs or special components such as baits and time-sorters (Woodcock, 2005;Brown & Matthews, 2016).
Data from the three sampling periods were pooled prior to analysis.The effect of funnel and roof on spider and carabid species richness and the abundances of functional groups were tested using mixed-effect general linear models (GLMM) with a poisson error term, typically used for count data and with a quasipoisson error term if we detected overdispersion in the data (Crawley, 2007).(R, lme4 package: version 1.1.12,glmer function; MASS package: version 7.3.23,glmmPQL function).The effect of roof (present versus absent), funnel (present versus absent) and their interaction (roof × funnel) were treated as the fi xed effects.To incorporate possible effect of spatial autocorrelation and differences in the structure of vegetation among sites, sampling site within habitat type (forest versus grassland) was used as the random effect.Separate models were run for forests and grasslands if the variability within a trap type was largely explained by the variance of the random effect.In this case, sampling site was the random effect.For goodness of fi t, marginal and conditional Rsquared values were estimated for the GLMM.
The statistical calculations were performed using the software R (version 3.3.2) (R Development Core Team, 2013).

RESULTS
In total, of 3279 spiders (2792 adults) and 2698 carabid beetles were collected, which belonged to 100 and 67 species, respectively (see Tables S2a-b and S3a-b).Distributions of the species and individuals caught by the different types of traps are given in Table 1.Furthermore, 45 vertebrates were caught, but pitfall traps with funnels caught fewer vertebrates: one amphibian and one mammal (see Table S4).
The species richness of spiders was not infl uenced by the type of trap (roof: z = 0.09, P = 0.929; funnel: z = -1.38,P = 0.166; funnel × roof: z = 0.86, P = 0.387, marginal R 2 = 0.028, conditional R 2 = 0.153).Furthermore, there was no (3) funnel pitfall trap without a roof and (4) funnel pitfall trap with a roof (Fig. 1).All traps were 500 ml white plastic cups; the roofs were made of white plastic plates held in position, ca.3-5 cm above the surface using 100 mm plastic sticks, and the funnels were cut out of 1500 ml transparent plastic bottles (PET, polyethylene terephthalate).The funnel pitfall traps also had an additional small transparent plastic cup for easier handling of the collected material and to prevent the funnel collapsing into the larger plastic cup.
We used 50% ethylene-glycol dissolved in water as a preservative and a few drops of odourless detergent to break the surface tension (Koivula et al., 2003;Schmidt et al., 2006).We established ten sampling sites in forest patches and ten in adjacent grassland.Four traps (one of each type) were placed at random in a quadrate at each site, with 8 m between the traps, this trap distance corresponds to the recommended minimum space between traps (Perner & Shueler, 2004;Zhao et al., 2013).As our aim was to directly compare the capture effi ciency of different types of traps we chose a relatively short distance in order to minimize the infl uence of differences in microhabitat conditions within sites (Lange et al., 2011).Although a small distance between traps may have decreased the numbers of invertebrates captured (e.g. Ward et al., 2001), it would, however, have had a similar effect on all the traps due to the random design.The closest sites were at least 50 m apart to reduce spatial autocorrelation between samples (Fig. 1).In total, we installed 80 traps, two habitats × ten sampling sites × four traps.Sampling was conducted in three periods in 2014, between May 5 and 12, July 24 and August 1, and October 9 and 20 (further information is given in Table S1).
Adult spiders and carabid beetles were identifi ed to species level according to Nentwig et al. (2014) and Freude et al. (2004) and were sorted into size classes in order to test the sampling ef-  difference in the total abundances of spiders or the abundance of the different functional groups (Tables 2 and S5).
The species richness of carabid beetles was also unaffected by type of trap (roof: z = -1.018,P = 0.309; funnel: z = -0.652,P = 0.515; funnel × roof: z = 1.141,P = 0.254, marg.R 2 = 0.007, con.R 2 = 0.561).The abundance of the carabid beetles caught was affected by the different types of traps (Tables 2 and S5).The total abundance of beetles was lower in funnel pitfall traps than in conventional traps.Roof alone had no effect, but the use of both roof and a funnel had a positive effect on the capture effi ciency.Type of trap had no effect on the abundance of small beetles, except for the positive effect of funnel pitfall traps with a roof on the capture of the S-F3 group.Regardless of their fl ight ability, fewer large beetles were collected by funnel pitfall traps than by the other types of traps.In the two habitats, the type of pitfall trap had a different effect on the abundance of large carabids caught (Tables 2 and S5).In forests, funnel pitfall traps collected fewer L-F2 and L-F3 carabids, and in grasslands, traps with a funnel and a roof had positive effect on the capture of the L-F3 group.

DISCUSSION
Numerous factors bias pitfall trap catches (reviewed by Brown & Matthews, 2016), however, this method has several favourable attributes such as low labour-requirement and simultaneous sampling at many locations.In this study, we compared the effects of different combinations of funnel and roof types of pitfall traps.In support of our hypothesis there was no bias in the capture effi ciency of arthropods by funnel pitfall traps with a roof in this study.We also showed that the suitability of different types of traps depends on the target taxa (spider vs carabids) and functional group (e.g.small vs large beetles).Type of trap did not affect signifi cantly the number of spiders caught, however, funnel pitfall traps collected fewer large beetles.Moreover, we show, that funnel pitfall traps with roofs catch fewer small vertebrates (Table S4).Funnel pitfall traps perform better than conventional pitfall traps for several reasons.They can be more effi cient in retaining invertebrates in the cups and thus collect more specimens (e.g.Vlijm et al., 1961;Baars, 1979;Obrist & Duelli, 1996), the use of a funnel prevents the evaporation of the preservative fl uid (Gurdebeke & Maelfait, 2002), and greatly reduces catches of vertebrates (Pearce et al., 2005;Lange et al., 2011;Brown &Matthews, 2016).However, in the present study we also found a negative effect on the number of large carabids caught.The diameter of the funnel neck was only 22 mm and the slope of the lateral wall of the trap was lower than in traps without funnels, allowing a greater number to escape (Cheli & Corley, 2010;Knapp & Ru zicka, 2012).
Our results are in accordance with Brown & Matthews (2016), as they emphasize the benefi ts of using rain guards.Roofs may reduce the dilution of the preservative caused by rain and the accumulation of litter in the traps.Litter that accumulates in traps without a roof may increase the chance of escaping, which presumably decreases the catches of arthropods.Roofs may also intercept active fl ying beetles, thus preventing their escape.In accordance with our results, Cheli & Corley (2010) also report no effect of roofs on the sampling of spiders.On the other hand, Siewers et al. (2014) report that the largest carabids and spiders were more frequently recorded in pitfall traps with plastic covers, than in pitfall traps with other types of cover, e.g., wire mesh.Colour of the roof has no effect on the capture effi ciency of carabid beetles and spiders (Buchholz & Hannig, 2009), however, opaque roofs may lower the sampling effi ciency of species of day-active carabids (Baars, 1979;Bell et al., 2014).
Considering the random effects, we suggest that habitat has an important effect on the catching effi ciency of Table 2.The effect of roof and/or funnel and their interaction on the numbers of spiders and carabids caught by pitfall traps.Conventional pitfall traps were used as a control.Regression coeffi cients (β), t and P values are given.N -number of individuals, S -small (<10 mm), L -large (>10 mm), F1 -non-fl ying, F2 -fl ying possible, F3 -fl ying, f -forests, g -grasslands.different types of traps, particularly in the case of large beetles.Presumably, the higher amount of litter in forests compared to mowed grasslands, where dead plant material is removed, may increase the chance of their escaping from traps without roofs.

Roof
Compared to the other combinations, funnel pitfall traps with a roof are the most effective method for sampling.The accidental catching of vertebrates can be avoided by using funnels, and a roof may indirectly results in reducing the number of invertebrates that escape, by preventing litter from falling into the trap and retaining beetles that try to fl y out of the trap.If the expectation is that a high number of large invertebrates will be caught, the diameter of the exit of the funnel has to be chosen carefully, especially if beetles are being sampled in a forest.Based on our results we emphasize that in addition to the sampling design, such as nested cross array (Perner & Schueler, 2004) or two-circle method (Zhao et al., 2013), the use of a suitable type of trap is also important for accurately estimating the density of ground-dwelling arthropods using pitfall traps.

Fig. 1 .
Fig. 1.Exploded view of a pitfall trap (a) and the layout of the sampling scheme (b).Conventional traps without a roof included only component d, conventional traps with a roof components d+a, funnel pitfall traps without a roof d + c + b, funnel pitfall traps with a roof d + c + b + a. Abbreviations of components: a -roof, b -funnel, c -small cup, d -large cup.

Table 1 .
The means (μ) and standard errors (SE) of the numbers of species and individuals caught by the different types traps.Abbreviations: conv -conventional, S -number of species, N -number of individuals, a -roof, b -funnel, c -small cup, d -large cup.

Table S1 .
Details of sampling.Table follows recommendation of Brown & Matthews (2016).

Table S2a .
The codes for the combined category of size and fl ying ability of the different species of ground beetles caught in the pitfall traps.S -small (<10 mm), L -large (>10 mm), F1 -non-fl ying, F2 -fl ying possible, F3 -fl ying.

Table S3a .
Numbers of each carabid species caught by each type of trap in each habitat.Abbreviations: g -grassland, f -forest.

Table S3b .
Numbers of each spider species caught by each type of trap in each habitat.Abbreviations: g -grassland, f -forest.

Table S4 .
Number of vertebrates caught in each type of trap in each habitat.Abbreviation: S -small.