Preliminary assessment of cavity‐nesting Hymenopterans in a low‐intensity agricultural landscape in Transylvania

Abstract In this study, our aim was to assess several traits of cavity‐nesting Hymenopteran taxa in a low‐intensity agricultural landscape in Transylvania. The study took place between May and August 2018 at eight study sites in the hilly mountainous central part of Romania, where the majority of the landscape is used for extensive farming or forestry. During the processing of the trap nest material, we recorded several traits regarding the nests of different cavity‐nesting Hymenopteran taxa and the spider prey found inside the nests of the spider‐hunting representatives of these taxa. We also evaluated the relationship between the edge density and proportion of low‐intensity agricultural areas surrounding the study sites and some of these traits. The majority of nests were built by the solitary wasp genus Trypoxylon, followed by the solitary wasp taxa Dipogon and Eumeninae. Solitary bees were much less common, with Hylaeus being the most abundant genus. In the nests of Trypoxylon, we mostly found spider prey from the family of Araneidae, followed by specimens from the families of Linyphiidae and Theridiidae. In the nests of Dipogon, we predominantly encountered spider prey from the family of Thomisidae. We found significant effects of low‐intensity agricultural areas for the genera of Auplopus, Megachile, Osmia, and the Thomisid prey of Dipogon. We also found that the spider prey of Trypoxylon was significantly more diverse at study sites with higher proportions of low‐intensity agricultural areas. Our results indicate that solitary bees seem to be more abundant in areas, where the influence of human activities is stronger, while solitary wasps seem to rather avoid these areas. Therefore, we suggest that future studies not only should put more effort into sampling in low‐intensity agricultural landscapes but also focus more on solitary wasp taxa, when sampling such an area.


| INTRODUC TI ON
Several recent studies have reported a decline of insect abundance, biomass, and species richness in many densely populated regions of Western Europe and also other parts of the world (Forister et al., 2019;Hallmann et al., 2017Hallmann et al., , 2021Sánchez-Bayo & Wyckhuys, 2019.  The two main drivers behind this decline are the increasing agricultural expansion and intensification as well as urbanization in these regions, which lead to a loss or fragmentation of the insects' habitats Knop, 2016;Merckx & Van Dyck, 2019;Piano et al., 2020;Raven & Wagner, 2021;Sánchez-Bayo & Wyckhuys, 2019Wagner, 2020). However, some other recent studies reported a partial recovery of insect abundance, biomass, and species richness in certain Western European regions (like in the Netherlands or Great Britain) since the 1990s, where different kinds of management actions or policies (e.g., stricter regulations of pesticide use, agri-environmental schemes, conservation programs) have been implemented to protect and maintain (insect) biodiversity (Carvalheiro et al., 2013;Ollerton et al., 2014).
It has already been demonstrated in numerous studies that trap nests are useful tools to assess the biodiversity of cavitynesting Hymenopterans and also their trophic interactions in a certain area as well as the parasitoids and hyperparasitoids of these Hymenopteran taxa (Albrecht et al., 2007;Klein et al., 2006;Kruess & Tscharntke, 2002;Mayr et al., 2020;Scherber et al., 2010;Staab et al., 2018;Stangler et al., 2015;Steckel et al., 2014;Tscharntke et al., 1998). Basically, cavity-nesting aculeate Hymenopterans can be divided into two trophic groups of nectar and pollen-feeding solitary bees and predatory solitary wasps (Klein et al., 2006;Mayr et al., 2020;Steckel et al., 2014). With regard to the pollination service provided by cavity-nesting solitary bees, which are pollinators of many wild and crop plant species, and the biological pest control by some cavity-nesting solitary wasp species (like Ancistrocerus gazella; Harris, 1994), additional knowledge about these species and the influence of landscape context on them may provide help in measures for their protection.
The fact that trap nests provide a good nesting opportunity and thus lead to an accumulation of cavity-nesting solitary Hymenopteran species living in the area surrounding these nests also makes trap nests especially suitable to study landscape effects. Some studies dealing with the effects of landscape context on cavity-nesting Hymenopterans conducted rather simple landscape analyses looking only at the presence of (Holzschuh et al., 2009;Mayr et al., 2020;Tscharntke et al., 1998) or distance from certain habitat types like forests (Klein et al., 2006) or ecological compensation area (ECA) meadows (Albrecht et al., 2007). Other studies, however, looked more specifically at the landscape structure surrounding their study sites, analyzing the effects of the proportion of different habitat types (Coudrain et al., 2016;Kratschmer et al., 2020;Taki et al., 2008) or even conducting complex landscape analyses (Holzschuh et al., 2010;Steckel et al., 2014) at multiple spatial scales (Steckel et al., 2014;Taki et al., 2008).
Most previous studies, which were assessing cavity-nesting Hymenopterans in different Western European countries, were conducted in high-intensity agricultural landscapes (Table 1). However, in the eastern part of Europe, there are still a few regions and areas remaining, which are not under such a strong human influence.
An example for such a region is Transylvania in the central part of Romania, where the population density is relatively low and the majority of the landscape is used for extensive farming or forestry. The most common form of extensive farming in this region is traditional small-scale farming, which is characterized by manual hay mowing, manual hay gathering, and extensive low-intensity organic manuring (Babai & Molnár, 2014;Babai et al., 2015). Such small-scale pastures and meadows often harbor a high species diversity of insects and are regarded as high nature value (HNV) grasslands (Veen et al., 2009), which are still widespread in the Transylvanian section of the Carpathian Mountains (Huband et al., 2010). Compared to Western Europe, however, there is a large gap of knowledge concerning the abundance and diversity of cavity-nesting Hymenopterans in Eastern Europe. Up to this date, only a few studies have taken on this topic in Eastern Europe (e.g., Budrys et al., 2010) and no study has addressed this issue in Transylvania. This highlights the need for more studies from such less-disturbed reference landscapes.
Therefore, the goals of our present pilot study were the following: (a) to assess and quantify the abundance and taxon diversity of the cavity-nesting Hymenopteran assemblage in our study area; |(b) to identify and quantify the spider taxa preyed by the spiderhunting representatives of the Hymenopteran taxa; (c) to analyze the influence of the proportion and edge density of low-intensity agricultural areas around the study sites on both Hymenopteran and spider prey taxa. Concerning our first goal, we were interested if we would encounter a different taxon composition of cavity-nesting Hymenopterans in the rural, low-intensity agricultural landscape of our study area compared to other, more intensively used Western European study areas (Table 1). Regarding our last goal, we were curious to find out which cavity-nesting Hymenopteran and spider prey taxa would be significantly affected by the proportion and edge density of low-intensity agricultural areas around our study sites.

| Study sites
The study took place in a hilly mountainous area at the border of the two counties Hargita and Kovászna (Transylvania, Romania), where the valleys are predominantly used for extensive, small-scale farming. The landscape surrounding our study sites can be defined as a cultural-historic low-intensity agricultural landscape, which consists of a mosaic of grassland and woodland patches. The grassland patches are mostly used as meadows and pastures, where grazing is made with low numbers of cattle and predominantly hand-mowing is applied ( Figure 1). The eight study sites were located in three valleys between 530 and 630 m a.s.l. ( Figure S1). The natural vegetation in this region at this sea level mostly consists of sessile oak-hornbeam or hornbeam-sessile oak (Querco petraeae-Carpinetum or Carpino-Quercetum petraeae) and hornbeam-beech or bastard balm-beech (Carpino-Fagetum or Melittio-Fagetum) mixed forests (Benke, 2004;Szabó, 1985). Two of these valleys were formed by the Vargyas creek (='Vargyas valleys') and are separated by a canyon ( Figure S1A). The third one is located 5-8 km east to the Vargyas valleys and was formed by the Körmös creek (='Körmös valley'; Figure S1B). The main flow direction of both creeks in this area is north to south. The Northern Vargyas valley is mostly used for extensive grazing and is dominated by meadows and pastures, while the Southern Vargyas valley, due to its remoteness, is much less used for grazing and more dominated by forest patches. Compared to the two Vargyas valleys, the Körmös valley is more strongly influenced by humans with arable land in its southern part, close to the settlement Erdőfüle (Filia). As a result of these differences in the intensity of land use, the ratio of low-intensity agricultural areas to the natural woodland and other areas in the close surroundings of the eight study sites also differed from site to site (Table S1). We established three sites each in the

| Trap nests
We installed four trap nests each at the eight study sites at the end of May 2018 ( Figure S1). All trap nests were marked with a unique code in reference to the sites and placed within 100 m distance around the center point. The trap nests were custom-made, consisting of a PVC tube of 12 cm diameter and 23 cm length ( Figure 2). The tubes were filled with stalks of common reed (Phragmites australis Cav.), which were cut off to a length of approx. 22 cm between the nodes, so that the inner part of the stalks would be freely accessible for any nest-building Hymenopteran. The stalks were placed tightly packed in the tubes to avoid them from falling out. The tubes were TA B L E 1 Examples for studies with a similar study design, analyzing the abundance and diversity of cavity-nesting Hymenopterans, carried out in different Western European countries placed in trees or shrubs at 1-2 m above ground. The trap nests were collected at the end of August 2018 and stored outdoors at a shady place. In January 2019, the nests were put into a refrigerator and stored at 4-7°C. In the same month, we began to collect the data from the reed stalks. For this, all stalks were cut open, and, in case we found a nest within a stalk, it was recorded with reference to the unique code of the trap nest plus a serial number, giving each nest a unique ID code. In case of each occupied stalk (=nest), we recorded the following parameters: (a) diameters of the reed stalks; (b) number of occupied brood cells, filled either with Hymenopteran offspring or spider prey (if present)-empty cells were also counted, but not used in further analyses; (c) type of nesting material; (d) color of larvae or cocoons (if present). Besides these parameters, we also counted the total number of stalks per trap nest. Based on the parameters (c) and (d), we were able to identify seven groups of nest types. From each of these seven groups, we also took a few nest samples (at least two) and reared them at room temperature in plastic bags. After the emergence of the adults from these samples, at least two specimens from each nest sample were collected, killed in 70% ethanol, and identified at genus level. We were able to identify the following eight genera: Ancistrocerus, Auplopus, Dipogon, Hylaeus, Megachile, Osmia, Symmorphus, and Trypoxylon. Except for the two genera Ancistrocerus and Symmorphus of the subfamily of Eumeninae (potter wasps), which could not be distinguished based on the nest type, each genus was assigned to a specific nest type. Therefore, based on this information, we distinguished between three taxa of solitary bees and four taxa of predatory, solitary wasps, giving them the name of the respective genus, except for the two genera of potter wasps, which were named after the subfamily.
If present, spider prey specimens were collected from the nests, put into 70% ethanol, and marked with the unique nest ID codes.
The spider prey were then taxonomically identified at species level-if possible, but at least at family level-and grouped according to the taxon of the spider-hunting wasp and the identified spider families.

| Landscape context
The landscape surrounding the eight study sites was mapped as land-

| Statistical analyses
All statistical analyses were conducted in R v3.6.3 (R Core Team, 2020), and all graphs were created using the R package 'ggplot2' (Wickham, 2016). The relationship between the nest numbers of solitary wasp and bee taxa was tested with a generalized linear model (GLM) assuming a Poisson distribution. We conducted principal component analyses (PCAs) using functions from the R packages 'FactoMineR' (Le et al., 2008) and 'factoextra' (Kassambara & Mundt, 2020). These PCAs were used to reveal if there was a relationship between the study sites and the nest F I G U R E 1 Typical landscape in the study area F I G U R E 2 A trap nest, mounted to a tree branch numbers of the Hymenopteran taxa and the specimen numbers of the most commonly preyed spider families, that is the families of Araneidae, Linyphiidae, and Theridiidae for Trypxylon (all above 100 specimens) as well as Thomisidae, which was the most frequently preyed spider family for Dipogon. All variables were scaled prior to the PCAs. The differences in the reed stalks' diameter used by the Hymenopteran taxa for nesting were tested with an ANOVA followed by a post hoc Tukey's HSD test (confidence level = 0.95).
The relationship between the number of nests and occupied brood cells for the seven cavity-nesting Hymenopteran taxa was tested with linear models (LMs).
We applied generalized linear mixed models (GLMMs) assuming a Poisson distribution from the R package 'lme4' (Bates et al., 2015) to analyze the effects of the proportion and edge density of the lowintensity agricultural areas on the cavity-nesting Hymenopteran taxa and the most commonly preyed spider families. In these GLMMs, The residuals of all LMs, GLMs, and GLMMs were tested for uniformity, dispersion, and outliers using functions from the R package 'DHARMa' (Hartig, 2020). We did not detect any significant deviations for the residuals of the tested models. Finally, we also checked for spatial autocorrelation (Moran's I) in the case of those data, where we encountered a significant effect of the landscape context, using the R package 'ape' (Paradis & Schliep, 2019). The coordinate reference system used for this analysis was ETRS89/ETRS-LAEA (EPSG: 3,035), the same one as used for mapping. We only detected significant spatial autocorrelation in the case of the brood cells of the genus Megachile (Table S2). Therefore, besides the normal linear regression models, we also used generalized least squares fits ('gls') by REML from the R package 'nlme' (Pinheiro et al., 2013), incorporating a Gaussian correlation structure in order to account for the spatial autocorrelation in case of Megachile. The brood cell numbers of Megachile were "log+1"-transformed for this analysis.

| Nests
In total, we found 990 nests in 4,857 reed stalks, with the occupancy per site ranging from ca. 13%-30% with a mean number of 20 ± 6% for all sites (  We partially found differences in the diameters of the reed stalks, which the Hymenopteran taxa used for nesting ( Figure 6).

| Spider prey
The largest number of identifiable spiders was preyed by Trypoxylon with a total number of 1,471 specimens (Table S5A), followed by Dipogon with 99 identifiable specimens (Table S5B)

| Low-intensity agricultural areas
The cell numbers of the Hymenopteran taxa of Auplopus, Megachile, and Osmia were significantly correlated with both the edge density and proportion of low-intensity agricultural areas around the study sites (Table 2). The strongest, significant relationships with low-intensity agricultural areas were found for Osmia, where an increasing edge density and proportion of these areas both had negative effects on the cell numbers of this bee genus ( Table 2). The cell numbers of the Auplopus wasp genus were significantly, positively correlated with an increasing edge density and negatively with an increasing proportion of low-intensity agricultural areas ( Table 2).
The cell numbers of the Megachile bee genus were significantly, positively correlated with both an increasing edge density as well as an increasing proportion of low-intensity agricultural areas (Table 2).
However, the effects of both the edge density (Estimate = 0.01; tvalue = 1.80; p-value = 0.12) and proportion of low-intensity agricultural areas (Estimate = 0.02; t-value = 0.60; p-value = 0.57) were not significant in the models corrected for spatial autocorrelation.
The number of Trypoxylon and Dipogon spider prey was largely unaffected by the edge density and proportion of low-intensity agricultural areas around the study sites (Table 3). From the preyed spider families, only the numbers of Thomisidae were significantly, negatively correlated with an increasing proportion of low-intensity agricultural areas (Table 3). The edge density and proportion of low-intensity agricultural areas had no significant effects on the SDI of the nest-building solitary Hymenopteran taxa at the study sites (Table 4). The SDI of the Trypoxylon spider prey, however, was significantly, positively influenced by the proportion of low-intensity agricultural areas around the study sites (Table 4).

| Nests
Analyzing the content of the trap nests revealed that the nest num-  Figure S1A and B). Thus, an increasing nest number of solitary bees might be the indication of an increasing human impact at the study sites.
Solitary bees were also the most abundant taxa in the majority of those Western European studies that were conducted in highintensity agricultural landscapes (Table 1), whereas in studies, that were carried out in natural (Sobek et al., 2009) or low-intensity agricultural (Albrecht et al., 2007;Krewenka et al., 2011;Kruess & Tscharntke, 2002)

| Spider prey
We found that the majority of spider specimens preyed by the genus Trypoxylon were from the family of Araneidae. In contrast to our findings, however, two other studies reported that the majority of spider specimens preyed by Trypoxylon figulus were from the family of Theridiidae (Coudrain et al., 2013;Hoffmann et al., 2020).
A possible explanation for the different findings of these two studies is that they were carried out in more intensively used agricultural landscapes. The results of the PCAs also support this assumption as they indicate that the Araneid prey of Trypoxylon was closely related to the study sites located in the remote Southern Vargyas valley (SV1-SV3), where the proportions of low-intensity agricultural areas were considerably lower than at the other study sites. However, the Theridiid prey of Trypoxylon was strongly associated with the study site K2, where low-intensity agricultural areas were the proportionally most dominant landscape element ( Figure 6).

TA B L E 2
Results of generalized linear mixed models (GLMMs) assuming a Poisson distribution, testing for the relationship between the proportion and edge density of low-intensity agricultural areas within 250 m around the eight study sites and the total number of occupied brood cells per nest and site, built by different cavity-nesting Hymenopteran taxa

| Low-intensity agricultural areas
The brood cell numbers of Osmia were significantly lower at study sites with both a higher edge density and proportion of low-intensity agricultural areas. This finding may come a bit unexpected, since most Osmia species feed on wild flowers, but many species are closely associated with forest habitats due to their nesting habits as they create small burrows for their nests in tree barks (Müller et al., 2019). In contrast to Osmia, a higher edge density and proportion of low-intensity agricultural areas both had a significantly positive effect on the brood cell numbers of Megachile. The brood cell numbers of the Pompilid were Auplopus were positively correlated with an increasing edge density, but negatively with an increasing proportion of low-intensity agricultural areas. This latter finding corresponds well with those reported by Holzschuh et al. (2009), who found that the abundance of Eumenid, Pompilid, and Sphecid wasps were highest at forest edges, which provide natural nesting sites, and lowest in grass strips, with a few natural nesting sites. They also reported that wasp abundance in grass strips connected to forest edges was higher than in slightly isolated grass strips and much higher than in highly isolated grass strips.
We did not detect any significant relationship between the edge density or proportion of low-intensity agricultural areas and the di- with the diversity being higher at study sites surrounded by a higher proportion of low-intensity agricultural areas. In other words, the lower the proportion of low-intensity agricultural areas was around the study sites, the higher was the proportion of Araneid specimens among the spiders preyed by Trypoxylon, which resulted in a lower diversity of Trypoxylon spider prey. The highest numbers of Araneid prey were encountered at the study sites SV1 and SV3, where the proportion of low-intensity agricultural areas was the lowest with regards to all eight study sites ( they also assumed that a higher proportion of grassland may cause Trypoxylon specifically hunting for its preferred prey species, resulting in a lower prey diversity found in their nests. that future studies not only should put more effort into sampling in reference landscapes with low-intensity agriculture but also focus more on solitary wasp taxa, when sampling such an area. As there are only a few such landscapes in Europe still remaining and as the maintenance of Hymenopteran biodiversity is crucial for the wellfunctioning of many ecosystem processes, our results can serve as a reference for future research in other areas, which are either less or more strongly influenced by humans.

CO N FLI C T O F I NTE R E S T
The authors of this article have no financial or other conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study were collected by the authors of this publication. The data that support the findings of this study are available in the Supplementary Information of this article. Additional data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.cjsxk sn64.

CO M PLI A N CE WITH E TH I C A L S TA N DA R DS
The manuscript is new in this form, it represents our own original work and has not been published, submitted or considered for publication elsewhere. The text, illustrations, and any other materials included in the manuscript do not infringe any existing copyright or other rights of anyone.