Introduction

Agriculture has shaped landscapes and influenced the behavior of wildlife over thousands of years. In the last century, intensification of agricultural land use led to increased field sizes and a reduced heterogeneity of farmland in many regions of the world. Additionally, an increased amount of chemicals and industrial fertilizers is applied by heavy machinery. These land-use changes have altered the prospects of many species across taxa, often leading to drastic declines of wildlife populations and biodiversity loss in agricultural landscapes (Foley et al. 2005; Lüscher et al. 2014; Stoate et al. 2001; Tscharntke et al. 2005) on a range of spatial scales (e.g., Matson et al. 1997; Wenzel et al. 2006).

In the year 2011, about 38% of the global land area was used by agriculture with about 30% of it covered by arable land (FAO 2015, date accessed: 05.11.2016; see also Additional file 1). In the face of a growing world population and the predicted agricultural land expansion and further intensification (Tilman et al. 2011; Tilman et al. 2001), it becomes increasingly important to understand the ecological dynamics in agricultural landscapes in order to mitigate environmental degradation and to sustain ecosystem functioning in the future.

In this context, European bats (Chiroptera) represent a specifically interesting taxon for two reasons. Firstly, the negative consequences of agricultural intensification, such as habitat loss, fragmentation, and reduced prey abundance, are known to affect bat populations (Dietz et al. 2007; Mickleburgh et al. 2002) and have led to severe population declines, resulting in the current level of protection in Europe (Council of the European Union 1992). Secondly, bats might act as biological pest control agents in agricultural landscapes (Maine and Boyles 2015; Puig-Montserrat et al. 2015) which highlights their value as a component of temperate zone agroecosystems. With the increasing perception of bats as ecosystem service providers in agroecosystems (Boyles et al. 2011; Kunz et al. 2011), researchers started to investigate a range of different spatiotemporal effects on bat activity in agricultural landscapes. On a local scale, prey availability, diversity, and land-use intensity affect the activity of bats above farmland (Wickramasinghe et al. 2003). On a landscape scale, several factors, such as specific landscape elements (Akasaka et al. 2012; Lentini et al. 2012), landscape composition, and configuration (Frey-Ehrenbold et al. 2013; Kalda et al. 2014) are known to affect bat activity and diversity above farmland. Hereby, the spatial scale of landscape characteristics which affect bat activity can differ between bat species (Akasaka et al. 2012; Lintott et al. 2016). Due to species-specific adaptations in wing morphology (Norberg and Rayner 1987), bat species differ in flight characteristics, such as flight speed and maneuverability. These flight characteristics are associated with the bat species’ mobility which can affect the spatial scale of habitat use (Bader et al. 2015). In addition, the habitat use of bats in agricultural landscapes changes with the season (Ciechanowski 2015; Heim et al. 2016) since it is tightly associated with the annual reproductive life cycle of most European bat species (Mackie and Racey 2007; Racey and Swift 1985).

Although previous studies investigated a large set of factors which are relevant for bat activity in agricultural landscapes, only a few included conventionally managed arable fields or focused on bat activity above arable fields (e.g., Ciechanowski 2015; Heim et al. 2016; Kelm et al. 2014; Lentini et al. 2012; Roeleke et al. 2016; Wickramasinghe et al. 2003). Additionally, studies that investigate not only specific factors but also their interactions are rare. For instance, by relating land-use intensity, insect abundance, and forest proximity to foraging attempts of bats, Treitler et al. (2016) found more foraging attempts on grasslands near forests, yet only on extensively managed grasslands. Furthermore, Heim et al. (2015) found that the forest extent in the surrounding landscape of grasslands can be more important for bat activity above grasslands during early summer compared to late summer. Thus, more studies which investigate the interactions of effects from different spatiotemporal scales are needed in order to improve our understanding of how and why bats use the agricultural landscape and which factors determine their movements, including their foraging activity. Such information could help in making conservation efforts for these bat species more efficient.

Here, we aim at identifying a set of spatiotemporal factors and their interactions which are relevant for species-specific bat activity above conventionally managed arable fields. We distinguished between relative bat activity and foraging activity, as bats may forage above fields but also cross them while commuting between roosts and foraging sites. We predicted that the relative activity of bats above arable fields will be largely affected by large-scale landscape characteristics (e.g., composition, configuration) while relative foraging activity will be largely affected by local characteristics (e.g., prey availability, crop type, vegetation height).

We hypothesized that the bat species’ mobility is associated with the spatial scale of landscape characteristics which affect bat activity. In particular, we predicted that the relative activity of highly mobile bat species will be explained best by landscape characteristics on a relatively large spatial scale compared to the activity of less mobile bat species which should predominantly respond to landscape characteristics on a relatively small spatial scale.

Methods

Study area

We conducted our study within the framework of the “Agricultural Landscape Laboratories” (AgroScapeLabs, www.bbib.org/scapelabs.html). The study area of the AgroScapeLabs covers about 291 km2 and is located in the Uckermark region (53° 20′ North, 13° 42′ East, Brandenburg, Germany, Fig. 1, for further details also refer to Heim et al. 2016).

Fig. 1
figure 1

Map showing the landscape composition of the study area (53° 20′ North, 13° 42′ East) (a) its location in Germany (b) and a close-up view of a site in a canola field with buffers in 300 and 500 m distance (c) (ArcGIS 10.0 © ESRI Inc. 1999–2010)

Full size image

This region is specifically interesting for our purpose as the amalgamation of small farms during socialist years (Behrens 2005) and the land-use change of the last 50 years resulted in large-scale changes creating large field units (20–75 ha; Katzschner 2011). These fields dominate the scenery and cover about 66% of the study area. Remnant forest patches (13%), grasslands (10%), water bodies (6%), and built-up areas (5%) represent a minor feature of the landscape. During summer, ambient temperature averages 17.4 ± 0.9 °C in Brandenburg (mean ± STD calculated based on data from 1981 to 2010 of the DWD 2016) with a precipitation of 567.1 ± 81.8 mm per year (mean ± STD calculated based on data from 1881 to 2016 of the DWD 2017).

Study design

The study design used here follows closely the design in the publication of Heim et al. (2016) as both studies were conducted simultaneously. We selected arable fields cultivated with corn (N = 18), canola (N = 18), and wheat (N = 17) since these were frequently used crop types in the study region. These crop types are economically important, as wheat and corn are two of the top five most produced crops in the European Union in 2014 (157 and 61 Mtpa), while canola was produced to a lesser amount of about 24 Mtpa (FAO 2015). From May to September 2012, we repeatedly assessed relative bat activity on a total of 53 arable fields (Fig. 1) by using a passive acoustic monitoring approach (Batcorder 500 kHz sample rate, 16 bits; EcoObs GmbH, Nuremberg, Germany). We aimed at monitoring each site once per month. In total, each field was sampled one to four times with a mean value of 2.5 times.

Batcorders were located in >150 m distance to landscape elements, such as forest edges, hedges, water bodies, and built-up areas to avoid recording bat activity which is influenced by an edge effect at such landscape elements (Heim et al. in prep.; Kelm et al. 2014). Recordings of bat activity were conducted only during nights with no or light wind and no rain. We randomly selected six sites of different crop types for monitoring on any given night and recorded bat activity simultaneously with one Batcorder per site within the first 3.5 h after sunset, which includes the first peak of nocturnal bat activity (Rydell et al. 1996). Hereby, the Batcorder thresholds were set to −36 dB and a critical frequency of 16 kHz. For further details, please refer to Heim et al. (2016).

Bat species identification

Out of the 24 bat species occurring in Germany, the following 13 bat species are known to occur regularly within the study region: Nyctalus noctula, Nyctalus leisleri, Eptesicus serotinus, Vespertilio murinus, Pipistrellus pipistrellus, Pipistrellus pygmaeus, Pipistrellus nathusii, Plecotus auritus, Barbastella barbastellus, Myotis myotis, Myotis nattereri, Myotis brandtii, and Myotis daubentonii (Teubner et al. 2008). We decided to identify bat species manually to avoid potential misidentifications of common bat species by automatic identification software (Russo and Voigt 2016; Rydell et al. 2017). The procedure of bat species identification closely follows the procedure in the publication of Heim et al. (2016). We measured start, peak, and end frequencies of at least two calls per recording using Avisoft SASLab Pro (version 5.1.0, Avisoft Bioacoustics, Berlin Germany; settings: Hamming window, FFT = 512, overlap = 93.75%; time resolution = 0.064 ms, frequency resolution = 0.977 kHz). To identify the bat species, we compared the call shape and the measured frequencies with published data (e.g., Miller and Degn 1981; Obrist et al. 2004; Parsons and Jones 2000; Pfalzer and Kusch 2003; Russo and Jones 2003; Rydell 1990; Skiba 2003; for details, see Additional file 2).

Relative insect abundance

For the assessment of the relative abundance and diversity of airborne insects, we caught flying arthropods at night using interception flight traps each equipped with a UV-light, a light sensor, and a plastic bottle filled with 70% ethanol attached to the cone of the trap. We trapped insects parallel to the recordings of the Batcorders. We used a distance of approximately 50 m to the recording site to avoid a bias of the insect trapping on the recorded bat activity, as bats might be attracted by the light of the insect trap. In the morning, insects were transferred into 99.8% ethanol for preservation. Later, we counted and categorized insects to order level using a stereomicroscope and identification keys (Köhler 2014). Additionally, we computed the Shannon Diversity Index per site on order level, based on the insect orders which occur in the study region. The insects which were caught on ground level (3 m above ground) might not represent the diet of bat species which forage at greater heights. Therefore, we included factors such as the local crop type, height, and vegetative status, which might affect insect abundance, diversity, or community composition on a larger scale, into our analysis. Thus, for each recording night and for each arable field, we estimated the crop height by measuring the height of several individual plants using a folding ruler (accuracy = ±5–10 mm) and documented the growth status of the crop based on two classes (1 = growing stage, empty field, or harvested; 2 = blooming and fruit building stage).

Landscape analysis

We used biotope and land-use vector maps of the federal districts of Brandenburg (BTLN 2009) and Mecklenburg-Western Pomerania (LUNG 2005). We reclassified landscape elements into six categories (see also Additional file 3): (i) agriculturally used areas (arable fields, grasslands); (ii) natural and semi-natural woody vegetation patches (forests, tree groups, tree lines, hedgerows); (iii) water bodies (≥1 ha, lakes, rivers); (iv) built-up area (buildings, sealed areas); (v) roads (streets, highways, railway); and (vii) kettle holes (<1 ha; remnant small water bodies). We used ArcGIS 10.0 (© ESRI Inc. 1999–2010) to create buffers of 300, 500, 1000, 3000, and 5000 m radii around the monitoring sites (Fig. 1). Hereby, the selection of the buffer zone ranges was based on the nightly flight distances from the roost reported for the bat species that occur in the study region (Dietz et al. 2007; Entwistle et al. 1996; Nicholls and Racey 2006a; Roeleke et al. 2016). With Patch Analyst 5.0.1.60 (Rempel et al. 2012), we analyzed different characteristics of the above-defined landscape elements. We assessed aspects of landscape composition by calculating the percentage of land covered [(Σ land use area/buffer area)*100%] by natural and semi-natural woody vegetation, water bodies, and built-up area patches within the respective buffer area to describe the amount of non-agriculturally used habitats. In order to describe aspects of landscape configuration, we assessed the shape complexity of natural and semi-natural woody vegetation patches [m/ha] by calculating the mean perimeter-area ratio for this habitat category. In addition, we included the density of kettle holes [ha−1] and the density of roads [m/ha] per buffer area, as kettle holes are very characteristic for this young glacial landscape and roads are known to influence bat activity on arable fields (Voigt and Kingston 2016; please see Additional file 4 for descriptive statistics).

Statistics

All statistical analyses were done in R (R Core Team 2014) and calculations of response variables are similar to the calculations in Heim et al. (2016). Following the one-zero time sampling approach (Martin and Bateson 1993), we divided a given night into 1-min intervals and counted the number of intervals with species-specific bat calls (Miller 2001) including both feeding buzzes and search calls. To obtain the measure relative bat activity for each species (hereafter: activity), we built a proportion based on the previously counted bat call intervals and the total number of 1-min intervals in a given night and site. To obtain the measure relative foraging activity for each species (hereafter: foraging activity), we counted the number of species-specific echolocation call sequences that indicate the pursuit of an insect, a so-called final buzz. Then, we counted the number of intervals with species-specific feeding buzzes and related this to the number of species-specific bat call intervals per night and site. Since both response variables are proportions, we used the binomial error distribution with the logit link in all following generalized linear mixed effects models (GLMMs).

Prior to model building, we tested for collinearity among explanatory variables using Spearman’s rank correlation and included only variables with ρ < 0.7 in the same model (Dormann et al. 2013). As we were not primarily interested in the effects of abiotic factors (temperature, relative humidity, cloud cover, wind speed) although they are known to affect bat activity (Ciechanowski et al. 2007), we only controlled for factors which were significantly correlated with the response variable within the respective models in a preceding analysis. As a result, we added the ambient air temperature to candidate models which explained activity and foraging activity of N. noctula (BAP: ρ = 0.27, p < 0.01; FAP: ρ = 0.34, p < 0.001) and P. pygmaeus (BAP: ρ = 0.26, p < 0.01; FAP: ρ = 0.22, p < 0.05), respectively. As seasonality is known to have a strong effect on bat activity in agricultural landscapes (Heim et al. 2016), we included Julian day as a covariate into all models. Furthermore, we log-transformed insect abundance, kettle hole density, and the percentage of water cover and of vegetation cover and scaled all variables before fitting GLMMs. Prior to model fitting, we explored the shape of the relationship between response and explanatory variables via generalized additive mixed models (mgcv::gamm4, Wood and Scheipl 2014). For that, we used the full model, a cubic regression basis and a random factor for recording site and fitted smoothing splines with four knots. Only the covariate “season” was associated with the activity of N. noctula, P. pipistrellus, and P. pygmaeus in a quadratic manner. Thus, we added a second order polynomial function of the covariate season to the respective set of candidate GLMMs.

The resulting 29 candidate models were based on our hypotheses and can be divided into four spatiotemporal factor sets based on local predictors (assessed on the scale of the site), landscape predictors across different spatial scales (landscape), temporal predictors (season) and their combinations, and the null model (Table 1).

Table 1 Candidate models are grouped into four factor sets and the null model: models with covariates related to local conditions at the site, models containing measures of landscape characteristics across five spatial scales, one model with a combination of local covariates, and models with various combinations of local and landscape-related covariates
Full size table

The first set contained models with local factors (crop type, height and status, insect abundance, and diversity), while the second set contained models with factors representing landscape characteristics only (Table 1). Hereby, variables in each section interacted with the season to identify whether effects change across seasons. In the third section, site-related variables were allowed to interact with each other and the season. In the fourth section of models, we combined landscape factors from each spatial scale with site-related factors and allowed them to interact with each other and with the season. This section contained also the global models with variables from all subsections combined (Table 1). For each species and response variable, we fitted all 29 GLMMs with the ID of the respective recording site as a random factor. We included an observation-level random effect where each data point received a unique level of a random effect, into the model of the activity of N. noctula to correct for overdispersion of the residuals (Harrison 2014). We checked the model fit by examining model residuals graphically using binned plots and tested for spatial autocorrelation (package ncf (Bjornstad 2016) based on Moran’s I). Hereby, we did not detect any signs of spatial autocorrelation in the residuals of the models. Then, we compared and ranked candidate models using the Akaike information criterion corrected for small sample sizes (AICc). By comparing all models to the null model, we tested whether any additional covariates explained more variance than the null model. By comparing the models from the first and second section to the third and fourth section, we tested whether and which combination of factor sets explains species-specific activity and foraging activity best or whether only landscape, local, or temporal effects are relevant. Furthermore, we were able to identify either one or several important spatial scales at which landscape characteristics explain bat and foraging activity best. If no single best model with an AICc weight (w i) of ≥0.9 could be identified after model ranking, we compiled a 90% confidence set of models (Burnham and Anderson 2002). In those cases where nested models occurred within the 90% confidence set of models, we selected the most parsimonious model for inference (for an overview of 90% confidence set of models, see Additional file 5). In a further step, we inspected the estimates of all covariates in the model and present only those covariates which showed an effect.

Results

Overall bat activity and comparison of overall activity between bat species

Out of 7409 bat recordings with at least two echolocation calls, we identified 6543 recordings (88.3%) to species level, 345 (4.7%) recordings to single-genus level (M. bra/dau, Pipistrellus sp., Myotis sp.), and 521 (7%) recordings to multi-genus level (Nyctaloid, unidentifiable; see also Additional file 6). Out of the 34,138 1-min intervals from 28 nights where a bat could have been recorded, only 3766 1-min intervals contained bat calls, which equaled an overall bat activity of about 11%. Out of these 3,766 1-min intervals, 12.4% were associated with feeding buzzes (Additional file 6).

Nyctalus noctula was the most frequently recorded bat species, making up about 35% of all recordings and occurring in about 85% of recording occasions (Additional file 6), while the three pipistrelles P. nathusii, P. pygmaeus, and P. pipistrellus were less often recorded (16, 11, and 6%, respectively). Although the three pipistrelles were less active than N. noctula, they occurred in about 55 to 80% of recording occasions (Additional file 6). With about 3 and 0.7%, respectively, M. bra/dau and M. nattereri were the least often recorded species (Additional file 6). Since GLMMs for these bat species did not converge or fitted badly, they had to be excluded from further analyses.

Local and landscape-scale effects

Bat activity was best explained by a combination of landscape characteristics and local factors (Table 2, for details on random effects, see Additional file 7). The only exception was the activity of P. pygmaeus, which was best explained exclusively by landscape characteristics and the season (Table 2, Additional file 8).

Table 2 Summary of the best generalized linear mixed effect models fitted to species-specific relative bat activity (BAP) and foraging activity (FAP) using the binomial error distribution. All 29 candidate models (see model no. in Table 1) were compared via AICc to identify the best model (w i ≥ 0.9; w i is the weight of evidence of a model given the set of models; based on Burnham and Anderson (2002)). If no best model was found, we used the 90% confidence set of models for inference by summing up models from top to bottom until the accumulated number of weights reached w i = 0.9 (Σ w i). Here, we present only the most parsimonious models, so that Σ w i values are smaller than 0.9 (complete overview in Additional file 5)
Full size table

We found that landscape characteristics often affected the correlation between insect abundance and bat activity (Fig. 2, see Additional file 9 for a tabular overview).

Fig. 2
figure 2

Estimates of factors derived from previously selected models which shape the relative activity of Nyctalus noctula, Pipistrellus nathusii, P. pipistrellus, and P. pygmaeus. Bat activity was modeled as a proportion of minutes with bat calls and the total number of minutes in a given night on a site (logit link). Interactions between landscape characteristics and either insect abundance (I) or season (S) are indicated by colons. Abbreviations: Kettle density density of small water bodies, Season 2 Julian day fitted using a second order polynomial function, S Additional file

Full size image

In particular, the direction of this correlation shifted depending on the condition of landscape characteristics (Fig. 3).

Fig. 3
figure 3

Effect plots for interaction effects show how low, medium, and high values of landscape characteristics affect the relationship between relative bat activity and insect abundance above arable fields. The spatial scale at which landscape characteristics were relevant is given in parentheses

Full size image

For example, P. pipistrellus activity decreased with increasing insect abundance above fields with a high proportion of woody vegetation cover and built-up area (Fig. 3a, b). On fields with a low proportion of woody vegetation cover and built-up area, P. pipistrellus bats were more active when insect abundance was high. In contrast to this pattern, we found that the activity of N. noctula, for instance, was positively correlated with insect abundance above fields surrounded by a relatively high density of roads (Fig. 3c). Similarly, P. pipistrellus bats were more active with increasing insect abundance above fields with a relatively high road and kettle hole density (Fig. 3d, e). In contrast, on fields with a relatively low road and kettle hole density, P. pipistrellus bats were less active with increasing insect abundance.

Spatial scale of landscape effects

In accordance with our expectations, the activity of N. noctula and P. pygmaeus was best explained by landscape characteristics at 5 and 1 km, respectively. However, the activity of P. pipistrellus was best explained by landscape characteristics on a 3-km scale, while the activity of P. nathusii was best explained by landscape characteristics on a 1-km scale (Table 2).

Degree of bat activity during summer depends on landscape characteristics

Interactions of landscape characteristics with the season occurred only in the models of N. noctula, P. nathusii, and P. pipistrellus (Fig. 2). In particular, depending on the parameter value of a landscape characteristic (low, medium, or high), bat activity above the arable field either varied across seasons or remained relatively high and independent of the season (Fig. 4, Additional file 10).

Fig. 4
figure 4

Effect plot shows how the strength of seasonal variation in the activity of Pipistrellus nathusii varies depending on the vegetation complexity (low, medium, high) within 1 km of the surrounding landscape

Full size image

For example, the activity of P. nathusii was lowest in June/July and increased until September above fields that were surrounded by vegetation patches with a relatively high degree of edge complexity (Fig. 4). However, the activity of P. nathusii did not differ between seasons above fields with a relatively low degree of vegetation complexity in the surrounding landscape. We found similar patterns in the interactive effects of water and vegetation cover with the season which influenced the activity of N. noctula and in the interactive effect of road density with the season which influenced the activity of P. pipistrellus (Additional file 10). Interestingly, all these patterns indicate that the changes in bat activity were strongest in summer, whereas activity in autumn remained on a relatively high and constant level.

Comparison of relevant effects across bat species

The activity of the three pipistrelles P. nathusii, P. pipistrellus, and P. pygmaeus was affected by similar landscape characteristics on similar spatial scales, however partially in opposite ways (Fig. 2). For instance, the amount of water cover and the density of kettle holes were associated positively with the activity of P. pygmaeus, while the kettle hole density was negatively associated with the activity of P. pipistrellus. Also, the activity of P. pipistrellus was positively associated with built-up areas, which were negatively related to the activity of P. pygmaeus. In addition, we found a positive effect of the percentage of woody vegetation cover on the activity of P. pipistrellus above farmland. Interestingly, the activity of P. nathusii was positively associated with the amount of water cover on the scale of 1 km which was also the case for P. pygmaeus and for N. noctula, however on a larger spatial scale and in interaction with the season.

Effects on species-specific foraging activity

The foraging activity of N. noctula and P. nathusii was best explained by a combination of factors related to prey availability and crop characteristics on the arable field and seasonal effects (Table 2, for information on random effects, see Additional file 7).

In general, crop height, insect diversity, and abundance on the sampled fields varied in the course of the recording period (Fig. 5).

Fig. 5
figure 5

Variation of crop height, insect diversity (on order level), and insect abundance from June to September 2012 on 53 arable fields where the insect and bat monitoring was conducted

Full size image

Overall, about 79% of the insects captured on all fields belonged to the order of Diptera (Additional file 11). The majority of the remaining insects belonged to the orders Coleoptera, Heteroptera, and Lepidoptera (about 10, 6, and 2%, respectively).

Foraging activity of both N. noctula and P. nathusii was positively affected by the crop height (Fig. 6), which interacted with the insect diversity in the model of N. noctula and with the insect abundance in the model of P. nathusii (Fig. 6, see Additional file 12 for tabular representation).

Fig. 6
figure 6

Estimates of effects on the relative foraging activity of Nyctalus noctula and Pipistrellus nathusii. The foraging activity was modeled as a proportion of feeding buzz minutes and the total number of minutes with bat activity in a given night and site (logit link). Estimates were taken from the best model selected by the Akaike information criterion. Note that all variables are related to the local scale of the sampled arable field. See also Fig. 7

Full size image

Hereby, N. noctula bats foraged most intensively at low to medium insect diversity values on fields with very tall crops (Fig. 7a). In contrast, P. nathusii foraged most intensively on fields with tall crops independent of whether insect abundance was high or low (Fig. 7b).

Fig. 7
figure 7

Effect plots for interaction effects show how low, medium, and high values of crop height influence the relationship between relative foraging activity and the proxies for prey availability

Full size image

Furthermore, we found that the foraging activity of N. noctula increased across months with the highest activity in September, whereas the foraging activity of P. nathusii was highest in June and decreased linearly until September. Interestingly, none of the used factors could explain the foraging activity of P. pygmaeus above fields, while the foraging activity of P. pipistrellus could not be modeled due to insufficient numbers of recorded feeding buzzes.

Discussion

We asked how European bat species use arable fields in a landscape dominated by conventional agriculture. To answer this question, we recorded the relative activity and foraging activity of bats at our study site in Germany which is dominated by intense agriculture.

We found a relatively low overall bat activity above arable fields compared to the bat activity above grasslands and in woodlands in a heterogeneously structured landscape only 50 km away from our study site (Heim et al. 2015; Jung et al. 2012; Treitler et al. 2016). The predominant bat species at our study site was N. noctula, followed by P. nathusii, P. pipistrellus, and P. pygmaeus. Bats belonging to the genus Myotis were least abundant. A combination of landscape scale, local, and seasonal effects explained the activity of most of the investigated bats.

In particular, we discovered that the influence of landscape characteristics could shift the correlation between bat activity and insect abundance into both positive and negative directions depending on the landscape characteristic condition, such as a high or low amount of water cover, and the type, such as whether it is water cover or road density. Furthermore, we found that the effect of landscape characteristics on the activity of N. noctula and P. nathusii above arable fields was associated with stronger variation in bat activity during summer compared to autumn. Finally, we found that the foraging activity of N. noctula and P. nathusii was explained by similar factors (crop height, insect related factors, and season). Yet, differences in the effects of relevant factors between bat species suggest that bats, although using the same foraging strategy of aerial hawking, might behave differently while foraging above arable fields.

Species-specific activity patterns

The observed differences in overall activity levels across bat species in this study were similar to those reported for grasslands embedded in a heterogeneously structured landscape about 50 km away from our study region (e.g., Heim et al. 2015; Jung et al. 2012; Treitler et al. 2016). The high abundance of N. noctula may best be explained by the numerous maternity roosts which are located in the study area. Furthermore, the differences in overall activity levels between bat species might reflect not only the abundance patterns but also their preferences for specific habitats (see Additional file 13 for details). Nyctalus noctula is well adapted to forage in the open space, while P. pipistrellus and P. pygmaeus use mostly the contact zone between arable fields and forests or hedgerows, the so-called edge-space (Heim et al. unpublished data). Myotis species are well adapted to forage in cluttered space and are therefore expected to avoid open space. Interestingly, although P. nathusii is categorized as an edge-space species, it appears to use the open space to a higher degree compared to the other two pipistrelle species (Kelm et al. 2014) and appears to be less dependent on linear woody landscape elements.

Landscape characteristics affect the relationship between bat activity and insect abundance

We found two distinct ways how landscape characteristics influence the relationship between bat activity and insect abundance on arable fields. Firstly, bat activity above arable fields was negatively correlated with insect abundance, if landscape characteristics with a positive effect on bat activity were present in the surrounding. For example, the activity of P. pipistrellus decreased with increasing insect abundance on arable fields with a relatively high amount of woody vegetation cover nearby, which seems contradicting. However, if we assume that the insect abundance measured on arable fields reflects an even higher abundance near, e.g., woody vegetation edges due to the accumulation of insects near such structures through wind (Lewis 1969), then our results indicate that P. pipistrellus might be drawn from arable fields to edges of woody vegetation in the surrounding landscape. This interpretation is confirmed by results obtained by Treitler et al. (2016) who found less feeding attempts of bats on grasslands further away from forests compared to grasslands close to forests. Secondly, we found that bat activity was positively correlated with insect abundance on arable fields if landscape characteristics were unfavorable in the surrounding. For instance, we observed that the activity of N. noctula and P. pipistrellus was positively correlated with insect abundance on the arable field if the surrounding was characterized by a relatively high road density and kettle hole density, respectively. Although N. noctula is expected to be least affected by the negative effects of roads due to its independence from linear landscape structures during navigation and its feeding high above the ground, it still can be found as roadkill on roads in other regions of the world (Altringham and Kerth 2016). Thus, the negative effect of road density on N. noctula found here highlights that this effect might be underestimated for N. noctula at least in landscapes dominated by agriculture. P. pipistrellus is not known to specifically select aquatic habitats such as kettle holes such as its sibling species P. pygmaeus (Lintott et al. 2016). Therefore, one possible explanation for the above-described positive relationship between N. noctula and P. pipistrellus activity and the insect abundance above the arable field might be that bats avoid unfavorable conditions associated with nearby landscape characteristics and instead make use of any elevated insect abundance above arable fields. These two patterns potentially point towards a trade-off between the foraging above arable fields, which might offer less prey, and the foraging in other habitats in the surrounding, which probably harbor more prey, but might be associated with increased competition or other risks.

In general, our results underline the flexibility with which bats efficiently use limited resources, such as patches of insects above arable fields. Hereby, the landscape composition and configuration appears to be of importance. Therefore, the potential provisioning of ecosystem services by bats above arable fields might also depend on landscape characteristics.

The level of bat activity above arable fields during summer depends on landscape characteristics

We found that depending on the parameter value (low, medium, or high) and type of landscape characteristics, bat activity either remained on a relatively high level across seasons or decreased to lower levels in summer, which represents the most energy demanding time period of reproduction (Fig. 4, Additional file 10). Our results indicate that landscape elements might be more important during summer than during autumn in defining bat activity. Largely motivated by movements between foraging grounds, bat activity during summer might be more strongly associated with landscape characteristics than bat movements during autumn which include dispersal of young bats and movements related to mating and migration. Our results indicate furthermore that the suitability of arable fields as foraging habitats for bats during summer might depend on the characteristics of the landscape. For instance, during summer, N. noctula bats were more active on arable fields with a relatively high water cover in the surrounding, while the activity of P. nathusii decreased with an increasing degree of vegetation complexity. To our knowledge, seasonal changes in the importance of landscape elements were rarely investigated. However, Heim et al. (2015) found that the effect of forest cover in a 200-m buffer around grasslands was more important for overall bat activity above grasslands during summer compared to autumn. Furthermore, Kelm et al. (2014) found an indication for such an effect in association with hedgerows in agricultural landscapes. Thus, our results confirm that the importance of landscape elements for bat activity—at least above open fields—can change across seasons. Additionally, this effect appears to represent a general pattern as it occurs in different bat species and is for each species, except P. pygmaeus, associated with different types of landscape features. Potentially, this effect is specifically important for bat species which use predominantly the open space for foraging or are migratory, which is both the case for N. noctula and P. nathusii.

Relative activity of insectivorous aerial-hawking bat species differs on three levels: habitat use, diet composition, and spatial scale

Based on the type of landscape characteristic which affected the activity of P. pipistrellus and P. pygmaeus and the direction of the effect, we assume that these two species differ in habitat use preferences. Our results are supported by past studies on the habitat use of these two cryptic bat species (Jung et al. 2012; Nicholls and Racey 2006b; Russ and Montgomery 2002; Vaughan et al. 1997). Even in urban landscapes, the proportion of freshwater had a similar effect on both species (Lintott et al. 2016). Thus, these two cryptic sibling bat species appear to differ in habitat use preferences across different landscape types.

In contrast, P. nathusii appears to overlap to some degree with the way both P. pygmaeus and N. noctula use arable fields. However, the differences in arable field use might be sufficiently high to allow bats to avoid potential competition. For example, our results on the foraging activity of N. noctula and P. nathusii indicate that both species might differ on the level of the foraging behavior above arable fields. In particular, P. nathusii might focus on the overall insect abundance, as these bats foraged most intensively at high insect abundance values irrespective of the crop height (Fig. 6). This interpretation is supported by our finding that the overall abundance of insects was dominated by the order of Diptera (Additional file 11), which was reported to play a major role in the diet of this bat species (Krüger et al. 2014; Smirnov and Vekhnik 2014; Vaughan 1997). In contrast, N. noctula foraged most intensively at low insect diversity values above arable fields with the tallest crops (Fig. 6). Interestingly, the lowest third of insect diversity (0–0.22 Shannon’s Diversity Index) on fields with the tallest third of crops (>214 cm) was largely associated with insects caught on corn fields in August. Under these conditions we caught slightly less Diptera, but relatively more Lepidoptera, Trichoptera, and Coleoptera (Additional file 14), which represent important prey items for N. noctula (Gloor et al. 1995; Kaňuch et al. 2005; Smirnov and Vekhnik 2014; Vaughan 1997). Therefore, we hypothesize that N. noctula might focus on specific insects that might be associated with corn. This is a very interesting topic which can serve as a starting point for future investigation, also in the context of ecosystem service.

In addition to differences in landscape characteristic types and foraging activity, also, the scale at which landscape characteristics affected bat activity differed across bat species. Our expectations regarding the relationship between the scale of landscape characteristics and the bat species’ mobility were not entirely met. In the study region, N. noctula covered an average distance of 15 to 27 km (Roeleke et al. 2016), while P. nathusii, P. pipistrellus, and P. pygmaeus covered a mean maximum flight range of 7 km (Dietz et al. 2007) and 2 km (Davidson-Watts and Jones 2006), respectively. Based on these values, our expectations were met only for N. noctula and P. pygmaeus. Although P. pygmaeus and P. pipistrellus should be similar to their flight ranges (Davidson-Watts and Jones 2006), the activity of P. pipistrellus was best explained by landscape characteristics on a larger spatial scale of 3 km as compared to P. pygmaeus. However, Nicholls and Racey (2006a) found that the home range size of individual P. pipistrellus was three times as large as the home range size of P. pygmaeus, which would support our findings on the one hand. On the other hand, the roosts of the radio-tracked P. pipistrellus and P. pygmaeus in the study of Nicholls and Racey (2006a) were located in very different landscape types. Thus, the home range size differences between both species might be a result of differences in a landscape specific resource distribution and not based on species-specific differences. Furthermore, the activity of P. nathusii was expected to be best explained by landscape characteristics on a relatively large spatial scale, as P. nathusii covers relatively large distances (Dietz et al. 2007). However, we found that the activity of P. nathusii was best explained by landscape characteristics on the spatial scale of 1 km, which is similar to the spatial scale obtained for P. pygmaeus. As our results do not completely match the predictions, we assume that the relationship between mobility and the spatial scale of relevant landscape characteristics might be influenced by additional factors such as the distribution of resources in the surroundings of the arable field, or the scale at which a species perceives these resources.

Conclusions

We gained novel insights on how insectivorous aerial-hawking European bat species use large conventionally managed arable fields in a landscape dominated by conventional agriculture. One of the key findings is that landscape characteristics of the surrounding farmland affect the relationship between bat activity and insect abundance. Thus, we would expect that changes in landscape structures around open fields translate into changes in the way bats use the open space above these areas.

The predicted increase in land-use intensity and area of arable fields in the near future could lead to further simplification of the landscape structure. Additionally, the insect abundance and especially the insect diversity could be further reduced. Already today, the low insect diversity due to the intensification of agricultural management is alarming (Benton et al. 2002; Biesmeijer et al. 2006). As a consequence of landscape structure, homogenization and reduction of arthropod diversity processes like natural pest control might get disrupted (Tscharntke et al. 2005) potentially leading to more frequent or severe pest insect outbreaks (Bianchi et al. 2006; Gardiner et al. 2009; Zhao et al. 2015). Therefore, it is important to increase the heterogeneity of agricultural landscapes in order to maintain and increase the control of pest insects by beneficial arthropods and bats.

In addition, we are still far away from understanding resource partitioning dynamics between bat species even in such a relatively simple landscape and bat ensemble structure, although we did observe potential patterns of resource partitioning in this study. Since the predicted continuation of land-use change might lead to increased competition among bat species through further changes in insect community structure and reduced area of refuge habitats, it becomes increasingly important to further investigate the interactions between species.

Based on our results, we recommend increasing the density of woody vegetation patches as well as promoting linear landscape elements to facilitate connectivity, which might be beneficial especially for edge-space bats and bat species from the Myotis group which were rarely recorded in this study. Furthermore, conservation measures should aim to preserve wetlands, such as lakes, rivers, and kettle holes, since these appear to be important areas for many aerial-hawking insectivorous bats. Finally, the large-scale increase in arthropod abundance and diversity should be the main aim of conservation efforts in intensively managed agricultural landscapes, as arthropods represent the fundamental source of energy not only for insectivorous bats but also for many other organisms in the agroecosystem.