The current study was conducted at five scree sites on limestone bedrock situated in different geomorphological units of the Western Carpathians, Slovakia (Fig. 1A, B). For more detailed characteristics of the study sites, see also the previous studies (Jureková et al. 2019, 2021):

This study included two sampling methods: soil extraction – (SS) and the subterranean pitfall trapping – (ST) (Fig. 1C).

Species richness, mean number of specimens (average of the three samples per given depth), and relative abundances (%) were presented as community characteristics to compare Collembola collected by the two different methods.

Spearman correlation analysis was used to test relations between the Collembola species richness of the two sampling techniques, and differences between them were analysed using the Wilcoxon Matched Pairs Test (Statistica for Windows, version 12, TIBCO Software Inc 2013).

Two graphs were used to express the number of species and their relative abundance separately covered by the soil samples, subterranean traps, and both methods.

Theoretical species richness was estimated for each site by diversity estimators from sample-based abundance data. By default, the biased corrected form of Chao1 along with log-linear 95% confidence intervals (CI) is used. For those datasets with a coefficient of variation of the abundance distribution greater than 0.5, the larger from the Chao1 classic and ACE richness estimators is recommended (Chao 1987; Chao et al. 2005; Colwell et al. 2012). Furthermore, the data were analysed using rarefaction procedures that are specifically designed to avoid the potential bias generated by uneven sampling. The estimation of species richness, rarefaction and extrapolation curves were calculated using the EstimateS software (Colwell 2013).

The similarities of Collembola communities with respect to two sampling techniques used were analysed using Non-metric multidimensional scaling (NMS) ordination based on species relative abundance/dominance (D > 10%). Autopilot with slow and thorough mode and Sörensen (Bray-Curtis) distance (recommend for community data) were selected. After randomization runs, a 3-dimensional solution was accepted as optimal. NMS analysis was performed by the PC-ORD 7 package (McCune and Grace 2002; McCune and Mefford 2016).

Vertical distribution of species richness and relative abundances of Collembola life forms across the scree profile were analysed using both collecting methods.

The relation between a species’ relative abundance and body length was evaluated in dominant species (D > 10%) (Tischler 1955). For determination of the species body length the maximum body size provided in the various literature sources was considered (e.g., Fjellberg 1998, 2007; Potapov 2001).

Based on the experience of the authors and data in the literature (Rusek 2007; Potapov et al. 2016), Collembola species were distinguished into four main life forms (see the individual characteristics in Rusek 2007) according to morphological and ecological adaptations to the soil environment:

The mean number of Collembola specimens collected by both methods at five scree sites was 3,818 (987 from soil samples and 2831 from subterranean traps), comprising totally 100 species, 79 collected from soil samples and 68 from subterranean traps (Table 1, Appendices 1–5).

Species richness and relative abundances of Collembola at the sites varied with respect to the sampling technique (Table 1, Appendices 1–5). With the exception of site ZA, a higher number of species was detected by soil samples compared to subterranean traps. In contrast, at all scree sites considerably higher relative abundances of species were recorded by subterranean traps than by soil samples.

Collembola species richness at the sites showed a non-significant Spearman correlation between the two sampling techniques (r = 0.36, P > 0.05, N = 20). Similarly, a high but non significant correlation, was observed for species richness at separate depths (n = 5): 5 cm (r = 0.72, P > 0.05), 35 cm (r = –0.30, P > 0.05), 65 cm (r = 0.20, P > 0.05) and 95 cm (r = –0.20, P > 0.05), although at a depth of 5 cm the correlation was strongly positive. The Wilcoxon Matched Pairs Test revealed non-significant differences in species richness detected by the two different sampling techniques (Z = 1.40, P > 0.05, N = 20).

Overall, about half of the species was shared by both sampling techniques, while 32% were found exclusively in soil samples and 21% in subterranean traps (Fig. 2). Graph illustrate the percentage share of species richness in the soil samples, subterranean traps, and the combination of both methods at the sites (Fig. 2A). Contrary to sites A and S, a higher proportion of species richness was observed shared by both methods compared to soil extraction and traps separately at sites B, ZA and ZB. Overall, a much larger number of species was captured by the extraction method, with the exception of sites S and ZA. Figure 2B demonstrates the significantly higher proportion of abundance shared by both methods at each study site compared to traps and soil samples, separately. With the exception of site B, significantly higher relative abundance was captured by the subterranean trap method compared with the soil samples.

The rarefaction curves plotting the number of individuals against the number of species for individual sites did not approach a horizontal asymptote (Fig. 3), indicating that during the study total species richness was underestimated in all cases. To get a more precise picture, species richness estimators (ACE for site ZA for soil samples, Chao1 for the rest) were generated for each site (Table 2). The greatest difference between the observed and estimated richness was at site A in the subterranean traps: 28 (CI: 18.7–37.3) and 83 (CI: 45.5–200.2) species, respectively; thus, only 33.8% of the estimated species richness was recorded in this case. Similarly, only about 45% of species were recorded at site ZB in soil samples. Surprisingly, a slight difference was observed between both parameters at sites B and ZA, both for subterranean traps, with 98% and 97% of the estimated species richness, respectively.

An NMS ordination diagram (Fig. 4) shows Collembola life forms at the sites and the collection method used. A three-dimensional solution was recommended by Autopilot and confirmed by the Monte Carlo permutation test, with a significance of P = 0.004 and a mean stress of 5.31 for real data and 250 runs for both real and randomised data. The best three-dimensional solution had a final stress of 4.08, P < 0.00001, after 78 iterations. Only species with a total dominance (relative abundance) greater than 10% are indicated in the diagram regarding three principal life forms. The diagram separated species into two well-defined clusters with the respect to the sampling method. The first cluster represented species collected by the pitfall traps. Epigeic Lepidocyrtus serbicus and euedaphic Deuteraphorura insubraria were associated with site A, hemiedaphic Pseudosinella thibaudi with site B, while euedaphic species Heteromurus nitidus dominated at both sites A and B. Euedaphic species Folsomia kerni, Kalaphorura carpenteri and Oncopodura crassicornis were associated with site S. Hemiedaphic Ceratophysella granulata and euedaphic Pygmarrhopalites pygmaeus were characteristic species at site ZA, while hemiedaphic Pygmarrhopalites principalis at ZB. Epigeic Lepidocyrtus lignorum, Plutomurus carpaticus and Pogonognathellus flavescens were characteristic species for the both sites ZA and ZB. The second cluster represented species extracted from the soil samples. Hemiedaphic Pseudosinella horaki was closely associated with site S, while hemiedaphic Folsomia penicula and euedaphic Isotomiella minor and Protaphorura armata were abundant at sites S and B. Hemiedaphic Folsomia quadrioculata and Parisotoma notabilis and euedaphic Proisotomodes bipunctatus were associated with site A. Hemiedaphic Folsomia manolachei and euedaphic Onychiuroides pseudogranulosus were characteristic for nearby sites ZA and ZB.

Vertical distribution of Collembola life forms along the scree slope profiles differed remarkably between the two methods (Fig. 5). In subterranean traps considerably higher species richness and relative abundances of atmobiotic and epigeic forms were captured at all sites compared to those extracted from the soil samples. Although some atmobiotic and epigeic species were also captured in subterranean traps deeper in the scree profile, their relative abundances were very low. These forms showed decreasing patterns of both community parameters with increasing depth. Moreover, a considerably high share of the relatively large hemi- and euedaphic species, such as Ceratophysella granulata, Heteromurus nitidus, Pygmarrhopalites principalis and P. pygmaeus, were recorded by traps in the middle and especially deeper scree layers at sites A, S and ZA. In soil samples, a higher abundance of hemi- and euedaphic forms was recorded compared to traps, showing a decreasing trend in abundance towards the scree depth. High species richness and relative abundance of small, less active hemi- and euedaphic forms, such as Folsomia manolachei, Isotomiella minor and Parisotoma notabilis, were recorded in the surface scree layer (soil) commonly at each site.

The relationship between species relative abundance and body length in the dominant collembolan species showed different trends regarding the two sampling methods (Fig. 6). In soil samples the abundance of Collembola had a decreasing trend with increasing body length. Small species, such as Proisotomodes bipunctatus, Folsomia manolachei, F. quadrioculata, Isotomiella minor and Parisotoma notabilis, were predominantly collected by this technique, with the exception of the large Protaphorura armata. In contrast, the abundance of species collected with subterranean traps had an increasing trend with larger body size. Medium and large species, e.g., Heteromurus nitidus, Lepidocyrtus lignorum, L. serbicus, Plutomurus carpaticus, Pogonognathellus flavescens and Pygmarrhopalites pygmaeus, were collected more frequently by traps; the only exception was the small species Oncopodura crassicornis.

In a given habitat, a combination of several collecting methods is required to obtain a reliable picture of such a diverse arthropod group as Collembola (e.g., Prasifka et al. 2007; Querner and Bruckner 2010). This study is the first attempt to assess the efficiency of recently popularized subterranean traps for invertebrate fauna occupying colluvial MSS biotopes with respect to the species richness of different life forms. Our results demonstrated that soil samples were more efficient for covering species richness at most MSS sites, whereas subterranean traps working during 6–7 months captured a substantial portion of the quantity, which is clearly the result of the long exposure time of these traps in this study. The soil extraction method thus appears to be a suitable complementary sampling method in addition to subterranean pitfall trapping, as already stated by other authors surveying soil Collembola in different habitats (Querner and Bruckner 2010; Querner et al. 2013; Nsengimana et al. 2017).

Species richness estimators and rarefaction curves are both traditionally used for comparing and assessing species diversity from sample units per site (e.g., Buddle et al. 2004; Querner and Bruckner 2010; Raschmanová et al. 2018). The rarefaction curves calculated from our data did not reach an asymptote at all sites and both methods, indicating that the species inventory was incomplete. Moreover, there was no obvious pattern in the obtained and estimated species diversity between sites and methods, which is probably associated with the low number of samples involved in these analyses. This result suggests that complementary sampling methods should be used and/or a greater number of traps should be installed to obtain a more complete picture of the community inhabiting MSS habitats.

Our study showed that the collecting method determines the captured species composition of the community. For example, small-sized euedaphic Proisotomodes bipunctatus, occupying a thermophilous talus habitat covered by mosses and tree vegetation near the Ardovská jaskyňa Cave, had only a random occurrence in subterranean traps. However, it was recorded as the most abundant species in the soil samples, preferably occupying upper scree layers that are consistent with its habitat requirements (Potapov 2001). Similarly, the medium-sized hemi- and euedaphic species Folsomia quadrioculata, Isotomiella minor and Parisotoma notabilis were abundant in the soil of the uppermost horizon at nearby sites A and S in the Slovenský kras Karst. These eurytopic species dwell in various types of habitats, such as pastures, meadows, thermophilous and also mountain forests (Potapov 2001; Fjellberg 2007; Raschmanová et al. 2016, 2018).

As already noted by some authors (e.g., Ivanov and Keiper 2009; Carneiro et al. 2016; Sommaggio et al. 2018), pitfall trapping usually overestimates large and motile species, and thus it does not provide an objective community pattern of ground-dwelling invertebrates. Our data pointed out that subterranean traps were effective in collecting not only surface-active (atmobiotic and epigeic) species, e.g., Lepidocyrtus lignorum, L. serbicus, Plutomurus carpaticus and Pogonognathellus flavescens, commonly documented as abundant species in traps from other MSS biotopes of the Carpathians Mts (e.g., Nitzu et al. 2014, 2018; Jureková et al. 2019), but this technique also covered larger and motile hemi- and euedaphic life forms. For instance, the large euedaphic Heteromurus nitidus with complete furca showed markedly high activity exclusively in pitfall traps along the entire scree profile at nearby sites A and S in the Slovenský kras Karst, and similarly considerably numerous in traps at site B in the Borinský kras Karst. However, this species was only incidental in the soil samples. In Central Europe, H. nitidus inhabits the soils of thermophilous forests and relatively warm caves rich in organic materials (Kováč et al. 2016). Likewise, the active euedaphobiont Deuteraphorura insubraria, that was exclusively collected by subterranean traps at thermophilous forest scree site A, is also known from beech forests with limestone outcrops at high elevations (Salmon et al. 2010).

It is obvious that the number of collembolan species inhabiting the interiors of forested screes and their activity decline with increasing depth and decreasing organic carbon content (e.g., Gers 1998; Pipan et al. 2010; Rendoš et al. 2016). In the present study, different trends in vertical stratification of Collembola life forms and their relative abundances were documented between both methods. Atmobiotic and epigeic species, forming greater part of the communities in pitfall traps compared to soil samples, were distributed along the entire scree profiles, but their abundance and species numbers had a rather strongly decreasing trend with depth.

High-gradient extraction of soil samples evidently supported species with small body size, whereas pitfall trapping advanced species with greater body size. Similarly, Querner and Bruckner (2010) compared the combination of soil sampling and pitfall trapping to collect collembolan assemblages in agricultural fields, with large surface-active forms as well as motile and large euedaphic species, e.g., Heteromurus nitidus, mostly caught by traps. Although Protaphorura armata is large and occupies deeper soil layers, it was abundant using soil sample extraction, which is in accordance with Querner et al. (2013). Thus, these two collection methods differ substantially in efficiency with respect to the body size and life forms of Collembola.

It was found that some obligate cave collembolans may occupy the deeper MSS layers, such as the small Neelus koseli (Rendoš et al. 2016, 2020), which in our study was captured exclusively by pitfall traps from the moist deeper layers of the upper site of the scree gully (ZB) in Zádielska tiesňava Valley, Slovenský kras Karst. In general, obligate cave invertebrates are captured by soil samples only very rarely (Raschmanová et al. 2018).

Finally, we must keep in mind that the 6–7-months timespan of pitfall trapping on one hand, and the simple collection of a soil sample at a given date on the other are difficult to compare in terms of vertical distribution of soil-scree Collembola. Moreover, there was almost a one year lag between soil samples collection and start of Collembola collection by subterranean traps.

The efficiency of soil sample extraction does not depend primarily on the fauna activity; therefore, this collecting method provides a relatively objective pattern of the actual spatial distribution of invertebrates in the soil profile at a given time (Yi et al. 2012; Tuf 2015). In contrast, comparing species inventories carried out by pitfall traps in different habitats is difficult, because capture efficiency is biased in many ways, i.e., sampling interval, degree of activity of individual taxa and their behavioural reaction to the conservation fluid (e.g., Woodcock 2005; Querner and Bruckner 2010; Knapp and Ružička 2012; Carneiro et al. 2016; Hohbein and Conway 2018). Furthermore, the vertical stratification of Collembola assemblages in the soil/scree may be markedly influenced by the microclimate across its depth profile (e.g., Hopkin 1997; Cassagne et al. 2003; Nitzu et al. 2014; Jureková et al. 2021). Regarding colluvial MSS, soil Collembola usually migrate to deeper scree levels with higher and more stable moisture during warm and dry periods (e.g., Nitzu et al. 2014; Mammola et al. 2017; Mammola 2019), which may elucidate great quantities of hemi- and euedaphic forms in pitfall traps at some sites.