The study was carried out in an agricultural landscape located in the southern Arava Valley, Israel ( Figure 1 ). The Arava Valley is a long and narrow valley (165 km long and 5-15 km wide), located in a hot and hyper-arid desert (Goldreich and Karni, 2001). Mean annual temperature is 25.4 °C (January: 15.8°C - July: 33.8°C). Average annual precipitation, which is mostly confined to a few rainy days in the winter (October-May), is 27.6 mm, with extreme interannual variation ¹ . Most agricultural production in that region is completely dependent on irrigation to grow seasonal vegetables (during Spring and Autumn). Sixteen species of insectivorous bats were observed within the agricultural ecosystems in the study area, and some were found to provide biocontrol services in the regional date plantations (Hackett et al., 2013; Schäckermann et al., 2022; Arzi et al., 2023). Most of the agriculture fields, in which we conducted our experiment and monitoring, are surrounded by planted windbreak barriers (singular tree line, genus Tamarix).

We conducted a field experiment and a field assessment in which we recorded bat activity above crop fields in relation to the broadcasting of calls or to windbreak hedgerows formation. As Bat activity, in general, varies considerably in both space (Kunz, 1973) and time (Hayes, 1997), we a) restricted our experiment to specific locations and temporal windows, and b) paired experimental plots with control plots, both in space and in time. First, as bat activity might be affected by moon phase (Appel et al., 2017), we limited the experiment to nights with less than 50% moonlight. Second, bat activity was always monitored simultaneously in paired plots (treatment and control). The plots were paired so that they exhibited similar structure in terms of vegetation (crop type and crop stage) and field spatial orientation (direction of the field and field edge structure) (Heim et al., 2018; Rodríguez-San Pedro et al., 2018).

The experiment was performed between October and November 2021 at 16 different sites that at the time of the experiment were planted with either cultivated pumpkin, Cucurbita pepo (3 sites), melon, Cucumis melo (6 sites), or onion, Allium cepa (7 sites). Pumpkin and Melon belong to the same family (Cucurbits) and share very similar pests in the study region (Dobrinin, 2020). Thus, we treated them as a single crop type (Cucurbits) in the experimental design and for any further analysis. In total, we conducted the manipulation experiment for 20 detector nights in onion fields, and 20 detector nights in Cucurbits fields. During each night of the experiment, we conducted the experiment in two sites that were at least 400 meters away from one another to minimize the possible influence of one set on the other. Within each site, we placed paired plots (a Broadcasting manipulation plot and a Control plot) in two corners of the same field that were separated by 200-300 meters ( Supplementary Figure S1 ). In the broadcasting manipulation plot, three ultrasonic speakers were positioned, mounted on tripods 1.5 meters above ground, angled 30° upward from the horizon and facing different directions. A bat detector was placed one meter behind the speakers and mounted on a tripod one meter above the ground, pointing 45° upward from the horizon ( Supplementary Figure S1 ). This setting of the bat detectors and speakers was designed to minimize the pseudo-recordings of the playback by the detector. The control plot included three tripods (without speakers) and a bat detector in the same setting as in the manipulation treatment.

On each experimental night, we monitored bat activity from sunset until 45 minutes before dawn. However, the broadcast started 100 minutes after sunset. This time was chosen according to the observed activity time of the two common tested bat species (see next section on calls recordings and broadcasting) in the study area (Schäckermann et al., 2022). Thus, each experimental night was divided distinctively into three phases: a pre-broadcasting phase of 100 minutes, a broadcasting phase of 160 minutes, and a post-broadcasting phase of 400 minutes ( Supplementary Figure S2 ).

We created a playlist consisting of full-echolocation call sequences of the two most active bat species in the study site (Hypsugo ariel and Eptesicus bottae, Schäckermann et al., 2022). The broadcasted echolocation sequences had been recorded at the study site with a Wildlife Acoustic Song Meter SM4, including an external ultrasonic microphone SMM-U1 (Wildlife acoustics, Inc., USA) at a sample rate of 96 kHz. Echolocation sequences which matched the clearness criteria (no sound distortion, overlapping of bat calls and loud background noises on the file) were edited with SASLAB PRO (Avisoft Bioacoustics, version 5.2.15, Germany) into short duration sequences (2.5-4 seconds), and background noises were removed by using high pass filtering (cutoffs: 40khz for H. ariel and 28khz for E. bottae). In total, the playlist consisted of 15 echolocation sequences of H. ariel and 9 sequences of E. bottae.

As mentioned above, the broadcasting phase in each experimental night lasted for 160 minutes. These 160 minutes were composed of 8 sections of 20 minutes. Each of these sections was composed of 10 minutes in which a unique combination of call sequence was broadcasted (‘broadcasting temporal unit’) and 10 minutes of silence (‘silence temporal unit’), which served as the local baseline (natural) bat activity ( Supplementary Figure S2 ). These playlists were loaded on two computers: Toshiba Satellite Pro R50-B with a Realtek high-definition audio soundcard (version: 6.0.1.7521) and Dell Latitude E5410 with an IDT high-definition audio CODEC soundcard (version: 6.10.0.6292). Both computers broadcasted the playlist via the ultrasonic loudspeaker AVISOFT USG player BL Pro 2 (Avisoft Bioacoustics, Germany) with a 16-bit resolution and a 96 kHz sampling rate, that was connected to two dynamic ultrasonic “vifa” speakers (Avisoft Bioacoustics, Germany). The speakers were powered by 3 external 12V, 7.2A acid batteries (ROSTEC Advanced Technologies Ltd, RC12-7S, Israel). The volume was adjusted manually each night to ensure maximum broadcast volume while avoiding sound distortions, as indicated by a clipping alert bulb on the speaker.

We monitored bat activity throughout the entire night across 12 different sites in the Arava Valley, sampling two sites each night. All sites were located at the margin of open fields used for growing seasonal vegetables and bordered by windbreaks. At the time of the monitoring, only 6 of these sites were cultivated with pumpkin, melon or onion. The windbreaks consisted of a single row of Tamarix spp. trees, approximately 10 meters tall. These linear single tree lines occasionally had gaps of 5-15 meters where agricultural dirt road intersected them. We placed a bat detector 10 meters away from the gap, facing the gap to monitor the activity within this area (“gap treatment”). A control detector was positioned 200 meters away along the same tree line as the gap, in a continues section without gaps ( Supplementary Figure S3 ). Both detectors were posited on the leeward side of the windbreak.

In both experiments, we monitored bat activity from sunset until 45 minutes before dawn. We sampled bat foraging activity using an acoustic bat detector (AnaBat SD2, Titley Electronics, Australia). Bat call analysis software (AnalookW, version 4.1z, www.hoarybat.com) was used to manually analyze the recorded calls, counting bat passes and feeding buzzes (FB). Calls were classified into species through visual examination of the sonogram based on species-specific acoustic characteristics known for the bats in the research area (Hackett et al., 2017). A sequence of calls was counted as a single bat pass if it met the two following criteria: a) it included at least two consecutive clear calls from the same species (Gillam, 2007), b) the gap between two consecutive calls in the sequence was shorter than half of the total length of the whole sequence (Song et al., 2019). Otherwise, the sequence was split and counted as separate bat passes. To eliminate the possibility of counting pseudo-recordings (i.e. recording of the broadcasted calls) as bat passes, we excluded from the analysis any recordings that had repeated bat sequences patterns that are typical of the composed playlists.

Throughout the broadcasting experiment, we recorded and identified 18,332 bat passes, with 98% identified to the species level. On average, 235.05 ± 25.08 bat passes were recorded each night. A total of 12 species were found active within the field crops, with four dominant species accounting for 86% of the total bat passes: Hypsugo ariel (46%), Rhinopoma cystops (24%), Tadarida teniotis (9%), and Eptesicus bottae (8%). Two of these species were mainly active either before (Rhinopoma cystops) or after (Tadarida teniotis) the manipulation phase of the experiment. Therefore, we didn’t analyze their activity in response to the manipulation. One species, Rhinolophus hipposideros, was extremely rare, with only two identified passes throughout the entire experiment ( Supplementary Table S1 ).

We found that bat activity (for all species combined) significantly decreased throughout the night ( Table 1 ). Additionally, there was a significant positive effect of average night temperature on bat activity ( Table 1 ). Species-specific analysis for H. ariel and E. bottae revealed a significant effect of crop type on the activity of E. bottae (GLMM, X² _df=1 = 11.72, P < 0.001). Eptesicus bottae exhibited a higher rate of bat passes in the cucurbit fields compared to the onion fields (5.95 ± 1.12, 2.24 ± 0.68, respectively per 60 minutes of activity, Figure 2 ). In contrast, H. ariel did not show a significant difference in activity between the two crops (15.4 ± 2.64, 13.5 ± 2.11 respectively, Figure 2 ).

Analysis of H. ariel activity across both cucurbit and onion fields revealed no significant difference in activity between the broadcasting manipulation and the control plots nor between the temporal units (broadcast and silent units, Figure 2 ). Analysis for the onion fields alone showed a significant interaction between time and plot type ( Table 1 ). At the beginning of broadcasting phase, H. ariel activity was higher in the control plot (the plot without actual broadcasting), but as the broadcasting continued, the manipulated plot exhibited higher activity than the control ( Figure 3 ). A similar pattern was observed also in the cucurbit fields analysis, though the trend was not significant. Additionally, there was a significant negative effects of wind speed on the activity of H. ariel ( Table 1 ).

Analysis of E. bottae activity in the cucurbit fields during the broadcasting phase revealed a significant decrease in activity in the manipulated plot compared to the control plot ( Table 1 , Figure 2 ). However, within each plot type, the activity level of E. bottae did not differ between the broadcasting and the silence temporal units ( Table 1 , Figure 2 ).

Throughout the Windbreak gap assessment, we recorded and identified 10,216 bat passes, belonging to 12 species (the same species assembly as in the broadcasting experiment). Additionally, 554 FB were recorded. Bat activity was significantly affected by the gap, with higher rates of bat passes and higher FB ratio at the gap compared to the continues tree line ( Table 2 ). On average, 477.25 ± 84.08 passes per night with a FB ratio of 0.07 ± 0.02 were recorded at the gap, compared to 374.08 ± 62.75 passes with a FB ratio of 0.02 ± 0.01 at the control ( Figure 4 ).

The application of broadcast manipulation on bats is known as a bat luring technique that increases capture rates at bat surveys (Hill and Greenaway, 2005; Loeb and Britzke, 2010; Braun De Torrez et al., 2017). In our study, we employed this method as a CBC strategy with the aim of directly enhancing the activity of desert dwelling insectivorous bat species, which prey on a variety of pests in the Arava Valley (Schäckermann et al., 2022). To our knowledge, this is the first time that a broadcasting manipulation (i.e. bat luring) method was tested in the context of pest control services (Tuneu-Corral et al., 2023). We found that the activity level of Eptesicus bottae was reduced in response to the broadcast ( Figure 2 ), while Hypsugo ariel showed no response ( Figure 2 ). The difference in the responses between the two species can be attributed to their different sociality levels and competition tolerance. Eptesicus bottae is known to roost and forge either solitary or in small groups of few individuals (Benda et al., 2006; Shehab et al., 2007; Korine, 2020). Further, studies on species from the same genus have demonstrated a deterring effect when using broadcasted social calls for luring (Quackenbush et al., 2016). Our results, shows a deterring effect at higher conspecifics activity levels (cucurbit fields) and no apparent response at lower activity levels (onion fields), therefore it is likely that E. bottae avoids relatively high bat densities. Alternatively, it is possible that the observed repellent response is due to ambient noise intolerance (von Frenckell and Barclay, 1987). As passing individuals may not perceive the broadcast as calls of a foraging bats but rather as ultra-sonics interfering sounds (i.e. masking) that can deter foraging bats (Masters and Raver, 1996; Gilmour et al., 2021). In contrast to E. bottae response, the activity of H. ariel was not affected by the broadcast, despite exhibiting higher conspecific baseline activity compared to E. bottae. The social structure of H. ariel has not been formally studied, but observations of over 200 individuals emerging together from a roost (Yom-Tov et al., 1992) and their tendency to forage in large groups over water (Kounitsky et al., 2015) suggest that this species, like others in its genus, is social. Socially foraging bats are more susceptible to bat luring (Quackenbush et al., 2016) and more prone to adopt eavesdropping behavior to facilitate foraging (Barclay, 1982), especially when food is scarce (Prat and Yovel, 2020). However, their response to broadcast manipulation have been shown to be influenced by both bat density and prey densities (Hügel et al., 2017; Lewanzik et al., 2019). For example, Pipistrellus spp. was attracted to broadcasted feeding buzzes when baseline conspecific activity was low, but was deterred by broadcasting as conspecific activity increased (Lewanzik et al., 2019). Racey and Swift (1985) observed that P. pipistrellus exhibited a shift in behavior, leaving the area or becoming aggressive towards conspecifics as the availability of prey relative to conspecific density decreases. In our study we did not find strong evidence that H. ariel was attracted to the broadcast. However, the significant interaction observed between plot type and time during the broadcasting phase suggests a change in response ( Table 1 , Figure 3 ). Our broadcast rate of activity (bat passes per minute of broadcasting) remained constant during the broadcasting phase. However, bat activity and food availability naturally fluctuate throughout the night (Hayes, 1997; Kuenzi and Morrison, 2003). If H. ariel exhibits a similar behavioral shift in response to conspecific density as described above for P. pipistrellus, then the impact of broadcasting foraging calls may vary depending on the local density of conspecifics. Notably, in our study, we broadcasted a mixed playlist of foraging calls of E. bottae and H. ariel. These two species belong to the same foraging guild (Korine and Pinshow, 2004) and share a similar diet (Feldman et al., 2000). Razgour et al. (2011) showed that interspecific competition among desert dwelling bats over water, in the Negev desert, affects bat community structure as well as foraging time. Therefore, it is impossible to determine whether the observed responses of each species were mainly due to a response to conspecific calls, heterospecific calls or any the combination of these. Interestingly, the activity of E. bottae was significantly lower in the onion fields compared to the cucurbit fields ( Figure 2 ). This may be due to the more homogenized structure of onion plants versus the complex structure of cucurbit plants, which could affect insect diversity. However, further research is needed to understand the drivers underlying this observation.

Our results indicate that bats utilize small gaps in windbreaks bordering crop fields for foraging. We recorded higher activity levels and more frequent prey capture attempts (FB) at the windbreak gap compared to continues windbreak sections ( Figure 4 ). These gaps, created by an agricultural road that intersects the windbreaks, seem to enhance bat activity and foraging. Bats typically avoid crossing both major and smaller roads, especially when these roads interrupt commuting routs, such as tree lines (Berthinussen and Altringham, 2012; Bennett and Zurcher, 2013; Medinas et al., 2019). However, our findings align with Foxley et al. (2023) which noted that agricultural roads might positively affect bat activity in agricultural landscape. We propose two explanations for this preference in foraging habitat. Firstly, the gap adds structural complexity to the windbreak, which facilitates the search and capture of potential prey by bats. In addition, the gaps may allow bats to better exploit and instantaneously track pest abundance at the two sides of the windbreak. Secondly, the wind flow patterns around windbreaks may play a role in the flight of the insects. Flying insects, particularly those considered ‘poor flyers,’ are often carried by winds and turbulence created around windbreak (Lewis and Dibley, 1970; Pasek, 1988). Further, gaps in windbreaks can act as wind funnels, concentrating wind flow (Bitog et al., 2012). Consequently, numerous flying insects may accumulate around the gap, potentially leading to higher prey density that bats can exploit. Additionally, the gap in the windbreak may compel insects to fly between tree canopies from one side of the gap to the other, exposing them to an open area where they might be more vulnerable to predation (Schroeder et al., 2009; Morris et al., 2010). It is important to note that we did not directly assess the abundance of flying insects (i.e., potential prey) in this study. Therefore, determining which explanation underlies the favorable foraging behavior at the gap requires further investigation.

The results from the direct CBC approach (broadcast manipulation) suggest that enhancing bat activity by broadcast manipulation may be feasible in specific scenarios. Our findings highlight the importance of the social behavior of the broadcasted bat species, while the inter-specific and intra-specific interactions may pose challenges to the method’s application for pest suppression. Additionally, echolocation calls are known to attenuate rapidly in the air (Lawrence and Simmons, 1982), making them audible to other bats only within short distances of 35-160 meters (Dechmann et al., 2009; Cvikel et al., 2015). Dechmann et al. (2009) observed a response to broadcast manipulation only within a few meters, whereas Loeb and Britzke (2010) reported higher bat capture rates but no increase in activity due to the broadcast. This also raises concerns regarding the effectiveness of using echolocation call broadcasts as a CBC practice. However, we suggest that a better fine tuning between the rate of the broadcasted echolocation calls with actual on-site bat activity may improve the outcomes. Furthermore, we suggest that adding an insect lurer (e.g. pheromone trap; Korine et al., 2022) at the broadcasting site may facilitate the enhancement of broadcast manipulations. Additionally, we recommend that future field experiments in broadcasting will broadcast the calls of a single species at a time.

The indirect approach (windbreak gap) demonstrated potential for enhancing bat activity and foraging. Our data suggest that small gaps within windbreaks provide a more favorable foraging habitat for bats in semi-arid agriculture environments. Which for some of the threatened desert bat species might be crucially important. These findings align with other studies that found a positive effect of landscape heterogeneity on biodiversity, insect abundance and bat activity (Tuneu-Corral et al., 2023). Further, Foxley et al. (2023) also reported improved bat activity near small roads, especially when at both sides of it there were trees or hedges. Therefore, we suggest that creating more frequent small gaps within windbreaks (which by themselves promotes bat activity) may facilitate foraging activity and improve pest control. Moreover, this method is relatively inexpensive and simple to implement before or after the windbreaks had been established. However, it is important to note that during the experiment, wind speed was relatively low (3.54 ± 0.57 m/s), and stronger winds may lead bats to prefer foraging at windbreak sheltered zone and avoid gaps (Verboom and Huitema, 2010). Our study did not account for crop damage from wind, nor did it consider pest and natural enemies’ abundances and dynamics. Therefore, further research on the long-term cumulative effects of small gaps at windbreaks, along with monitoring their impact on crop yield, bat activity and pest’s abundance is needed. Nevertheless, reducing pesticides use in the Arava area by adopting CBC practices could support and benefit local bat populations and control the pest populations.