High detectability with low impact: Optimizing large PIT tracking systems for cave‐dwelling bats

Abstract Passive integrated transponder (PIT) tag technology permits the “resighting” of animals tagged for ecological research without the need for physical re‐trapping. Whilst this is effective if animals pass within centimeters of tag readers, short‐distance detection capabilities have prevented the use of this technology with many species. To address this problem, we optimized a large (15 m long) flexible antenna system to provide a c. 8 m2 vertical detection plane for detecting animals in flight. We installed antennas at two roosting caves, including the primary maternity cave, of the critically endangered southern bent‐winged bat (Miniopterus orianae bassanii) in south‐eastern Australia. Testing of these systems indicated PIT‐tags could be detected up to 105 cm either side of the antenna plane. Over the course of a three‐year study, we subcutaneously PIT‐tagged 2,966 bats and logged over 1.4 million unique detections, with 97% of tagged bats detected at least once. The probability of encountering a tagged bat decreased with increasing environmental “noise” (unwanted signal) perceived by the system. During the study, we mitigated initial high noise levels by earthing both systems, which contributed to an increase in daily detection probability (based on the proportion of individuals known to be alive that were detected each day) from <0.2 (noise level ≥30%) to 0.7–0.8 (noise level 5%–15%). Conditional on a low (5%) noise level, model‐based estimates of daily encounter probability were highest (>0.8) during peak breeding season when both female and male southern bent‐winged bats congregate at the maternity cave. In this paper, we detail the methods employed and make methodological recommendations for future wildlife research using large antennas, including earthing systems as standard protocol and quantifying noise metrics as a covariate influencing the probability of detection in subsequent analyses. Our results demonstrate that large PIT antennas can be used successfully to detect small volant species, extending the scope of PIT technology and enabling a much broader range of wildlife species to be studied using this approach.

Therefore, it is critical to improve these techniques to enable effective research approaches without significantly impacting the bats' viability.
Banding has been used to mark bats since 1910s (Allen, 1921), but can cause significant injury and lower survival in some species (Baker et al., 2001). Another alternative is radio-tracking; however, a comprehensive review has found that most radio-tracking devices used to study bats are too heavy, are being used with minimal ethical justification, and remain attached for an average of just 9 days (O'Mara et al., 2014). A more recent innovation for the use on small bats has been miniaturized GPS tags; however, currently this has only been successfully attempted with the use of anesthesia and sutures to attach the loggers-and battery life, tag weight and recapture rates remain ongoing issues (Castle, Weller, Cryan, Hein, & Schirmacher, 2015;Weller et al., 2016).
An alternative to these marking and tracking methods are passive integrated transponder (PIT) tags, which weigh as little as 0.1 g, which is well under the 5% of body mass "rule" recommended for bats under 70 g (Aldridge & Brigham, 1988;Neubaum, Neubaum, Ellison, & O'Shea, 2005). To date, PIT-tags have shown no apparent effect on body condition or reproductive success of small bats (Rigby, Aegerter, Brash, & Altringham, 2012).
PIT-tags are glass-encapsulated microchips that are injected into an animal and lay dormant until they are activated by a hand scanner or antenna system, which reads the tag's globally unique identification number using radio-frequency identification (RFID).
By positioning antenna systems at key locations, individuals can be passively tracked for a lifetime with just a single trapping event.
PIT-tag technology has been used extensively in fish research since 1980s and has also been used to study birds, reptiles, amphibians, invertebrates, and mammals (Gibbons & Andrews, 2004;Schlicht & Kempenaers, 2018;Soanes, Vesk, & Ree, 2015;Unger, Burgmeier, & Williams, 2012). PIT-tag technology has advanced the study of movement patterns and survival of wildlife; however, a major limitation of this technology has been low detection distance, with tagged individuals normally needing to pass within 30 cm or less of an antenna to be detected (Adams & Ammerman, 2015;Gibbons & Andrews, 2004;Norquay & Willis, 2014).
To date, microbat studies using PIT-tags and passive detection have been limited to close-range applications, typically using loop antennas at small roost entrances, such as tree hollows (Garroway & Broders, 2007;O'Donnell, Edmonds, & Hoare, 2011;Toth, Dennis, Pattemore, & Parsons, 2015), bat boxes (Godinho, Lumsden, Coulson, & Griffiths, 2015;Kerth & König, 1996;Kerth & Reckardt, 2003), or small building entrances  O' Shea et al., 2010;Safi, König, & Kerth, 2007). Bats have also successfully been detected at an artificial water source when tagged individuals came within 15 cm of a submerged plate antenna (Adams & Hayes, 2008). Large roost entrances, such as with caves, provide additional challenges for using PIT antennas due to the detection ranges required. PIT antennas have been installed on timber frames partly covered with mesh to modify the cave exit and funnel the bats through small "windows" (Britzke, Gumbert, & Hohmann, 2014), or by using a serpentine antenna configuration that zig-zags across the cave entrance (Adams & Ammerman, 2015). A drawback of these approaches is that they altered the flight path of the bats and led to short-term responses and effects such as circling, landing on the infrastructure, avoidance behavior, and wing-strikes. Furthermore, some bats have limited tolerance to structural changes at their roost entrances. For example, gates at caves and mines have caused some bats to modify their behavior and have been linked to declines in numbers and, in some cases, total site abandonment (Pugh & Altringham, 2005;Slade & Law, 2008;Tuttle, 1979). Increased predation risk can also result with predators using infrastructure to catch bats exiting the roost (White & Seginak, 1987).
The critically endangered southern bent-winged bat (Miniopterus orianae bassanii) is an obligate cave-dwelling bat with a restricted distribution in south-eastern Australia. The national recovery plan for the southern bent-winged bat recommends investigating and developing techniques that would enable PIT technology to be used for quantifying age-and sex-structured survival rates and to help identify the cause of population decline (Lumsden & Jemison, 2015), whilst minimizing trapping occasions and disturbance. Bent-winged bats in south-eastern Australia generally favor large caves with relatively large entrances (Dwyer, 1963) and do not readily accept cave gates (Slade & Law, 2008;Thomson, 2002). Therefore, modifying cave entrances to detect PIT-tagged southern bent-winged bats were deemed to pose an unacceptable risk to the species. As a result, there was a need to develop a system that could detect bats as they flew through large passages, without impacting their behavior.
Here we describe the challenges and successes of using large PIT antenna configurations for monitoring a small, volant, and fastmoving organism, the southern bent-winged bat, over a three-year period. We PIT-tagged and monitored 2,966 individuals, optimized an RFID system to successfully meet our aims of high detectability and low impact, and provide recommendations for other researchers considering using this technology for other wildlife species.

| Study sites
Our study was based at two limestone caves used by the southern bent-winged bat in south-east South Australia. From spring to autumn, southern bent-winged bats form a large colony at their primary maternity cave, Bat Cave, in the Naracoorte Caves National Park, World Heritage Area. This was our primary study site. It is a horizontal cave system with a roof window entrance measuring approximately 7 m by 4 m. Mains (240 V AC) electricity is connected to the cave to power permanent infrared and thermal cameras located inside the cave. These cameras transmit live images of the bats to the nearby Bat Observation Centre for visitor tours which form part of the tourist attractions for the national park. The secondary study site was a nonbreeding cave located on private property near Glencoe, South Australia, about 72 km from Bat Cave. The entrance of this cave measures approximately 6 m wide by 2 m high and is fenced from livestock. The southern bent-winged bat is the only bat species that is known to roost in these caves.

| RFID systems and installation
The radio-frequency identification (RFID) system used in this study was the Biomark IS1001-a low frequency (134.2 kHz) system with dynamic, automatic tuning. The system consists of a reader, data logger, and 15-m flexible cord antenna that collectively powers, detects, and records PIT-tags. The antenna is intended to be configured as a loop that detects tagged individuals as they pass through the loop. All data are recorded as log files to the internal memory or a USB flash drive connected to the data logger. We coupled this system with Biomark high performance 12.5 mm FDX-B tags (HPT 12) which are reported by the manufacturer to provide a greater readrange than other PIT-tags of the same size.
The entrance of Bat Cave was too large for the 15 m antenna. The narrowest section of the cave (hereafter, referred to as "the restriction") is located approximately 100 m from the entrance, measures approximately 5 m wide and up to 2.8 m high, and was identified as the most suitable position for the RFID system. Despite the distance from the cave entrance, we assumed that bats would fly through the restriction when present at Bat Cave because the restriction is a high traffic area for bat movement. Individuals fly through this passage to access the three main roosting chambers, including the maternity chamber where the bats raise their young (Dwyer & Hamilton-Smith, 1965). Individuals also move through the restriction during the day to drink from dripping stalactites in the "drinking chamber", located in an alcove off the main passage between the entrance and the maternity chamber (Codd, Clark, & Sanderson, 1999).
The placement of the antenna at the restriction in Bat Cave needed to satisfy two conditions: firstly, that bats would not collide with the antenna nor have their flight paths altered, and secondly, that the antenna configuration gave the best coverage and sensitivity for reading the tags. Despite the length and flexibility of the antenna, the Biomark IS1001 cord system will not successfully create a detection field in all configurations. Large rectangular antenna configurations (with a width exceeding 2.5 m) are likely to obtain the greatest antenna sensitivity by minimizing the height of the rectangle as much as possible (ideally to 1 m) and by laying the excess antenna cable close together (K. Pomorin, Karl Tek, pers. comm.).
As such, the dimensions of the restriction at Bat Cave were substantially larger than those advised for successful PIT-tag detection.
To determine the optimal antenna placement and configuration, we first observed the flight path of the bats through the restriction for several hours (including during a dusk fly-out) in August 2015, using a thermal camera (FLIR Photon 320). Analysis of the footage demonstrated that the bats flew in the upper-half of the restriction. It was therefore determined that the cord antenna could be safely configured with the bottom of the antenna setup to 1 m above the cave floor, thereby creating a more desirable height for the rectangular antenna configuration. As metal can interfere with antenna performance (Biomark Inc, 2015;Freeland & Fry, 1995), the cord antenna was attached to the wall and ceiling of the cave using plastic saddle clips drilled into the limestone. The bottom of the antenna was supported off the ground with flexible fiberglass poles which were drilled into the cave floor and fastened using plastic cable ties. The final dimensions of the antenna were a maximum of 4.8 m wide and 1.8 m high (Figure 1). The excess cord of the antenna was laid together in parallel (touching, or close to touching) and kept in place with cable ties.
The system was connected in January 2016 and powered using two battery banks because the Biomark IS1001 RFID system is not compatible with Australian 240 V AC mains power. Each battery bank was comprised of two deep-cycle 12 V DC batteries run in series to create an output of 24 V DC. The batteries were charged by a battery charger (CTEK MXT 14) connected to the mains power supply in the cave. The charger and batteries were separated from the RFID system by a battery-switcher unit (Biomark standard battery-switcher) that switches between charging and drawing power from each of the battery banks on a three-hour rotation. This system is designed to sustain the life of the batteries by ensuring the batteries are not drawn too low and to keep the RFID system electrically isolated from mains power so that it did not interfere with the system's performance. The batteries, charger, and battery-switcher were placed in plastic tubs with ventilation holes.
After the antenna was installed, the restriction was monitored

| Trapping and tagging
We trapped and PIT-tagged bats at Bat Cave in January and February over three consecutive years, 2016-2018. Bats were trapped with Austbat harp traps (Faunatech, Mount Taylor) set exterior to the fence that surrounds the cave entrance. Each bat was PIT-tagged using a sterilized 12-gauge needle (Biomark MK10 implanter and N125 needles in 2016, and Biomark MK 25 Implant Guns and HPT12 Pre-load Trays in 2017 and 2018). The 12-mm tag (Biomark HPT 12) was injected subcutaneously so that the tag rested between the scapulae, and the injection site was sealed with a drop of surgical glue (3M™ VetBond™) to minimize tag loss (Lebl & Ruf, 2010). A total of 2,966 southern bent-winged bats were tagged over the course of the study (approximately 1,000 per year). During the handling and tagging, the bats typically remained calm and were released minutes after the procedure (van Harten et al., 2019). All trapping, handling, tagging, and data collection procedures were approved by the La Trobe Animal Ethics Committee (AEC15-67) and the South Australian Department of Environment and Water (U26453).

| Data collection
The Biomark IS1001 data logger recorded two types of data chronologically into daily log files: tag data and system data. Each tag detection is recorded with the exact time and date of detection and the PIT-tag's unique identification number. The system data records status and noise reports which include system settings and noise levels (i.e., unwanted signal). As a range of system settings can be chosen, the settings used in this study are provided in Appendix 1. Full status reports were generated by the system hourly and noise reports were recorded every five minutes. Data files were recorded directly to USB flash drives plugged into the data logger board. Data were collected from the study sites regularly (approximately monthly) by manually retrieving the flash drives. Other system maintenance included initiating a full tune of the antenna using the BioTerm program (Biomark) on a laptop connected to the RFID system via the mini USB port (undertaken approximately every two months) and the installation of a software update to both of the Biomark IS1001 units (undertaken once).

| Quantifying and minimizing noise
Noise is the summation of unwanted in-band frequency signals being received by the RFID system, including electromagnetic interference and natural environmental factors, which degrade system performance by competing with the tag signal. The Biomark IS1001 measures noise as "FDX-B signal" in millivolts (0-900 mV range) and then converts this measurement into a percentage for ease of reference. At Bat Cave, initial noise levels were high (>25%). Potential sources for electromagnetic interference included the five pre-existing thermal and infrared cameras situated in various chambers of the cave that were linked to the Bat Observation Centre. Associated with the cameras was a network of 240 V AC cabling. To find and eliminate the source/s of the noise, we turned off power to the cave, measured noise levels (by initiating a noise report with the BioTerm program), and then systematically turned back on each of the cameras and cabling networks. After each change, the read-range and noise levels were recorded.
F I G U R E 1 Southern bent-winged bats flying through the 15 m loop antenna which was installed at the restriction in Bat Cave. The bottom of the antenna loop was raised above the cave floor. Only a small area of the restriction (to the right of the stalagmites) was not included in the detection space. Note that the "tail" of excess antenna cord was laid together and leads to the RFID reader on the right-hand side of the image. Boxes containing the batteries, charger, and battery-switching unit are located to the right of the camera's field-of-view. The structure in the middle of the photograph is a decommissioned infrared camera which provided real-time footage to the Bat Observation Centre prior to our study-other cameras are still in operation in other parts of the cave To decrease noise levels and increase system performance, we electronically earthed the RFID system at Bat Cave on 4 May 2016.
The floor of the cave near the antenna is bed rock, with little to no available earth or soil. Dry limestone is a poor electrical conductor, so instead of drilling and inserting an earth rod into the floor of the cave, we buried a two-meter copper earth rod horizontally under a thin bed of bat guano. The earth was connected to the exposed negative post-terminal of the Biomark IS1001 with a saddle, 6 mm earthing cable and ring terminal. Two additional rods were attached in series with additional saddles and earthing cable in February 2017 in an attempt to strengthen the earth. An earth was also added to the system at Glencoe on 7 May 2018, by hammering a 1 m long earthing rod vertically into the soil near the entrance of the cave and then connecting the rod to the negative terminal of the Biomark IS1001.

| Read-range and the impact of noise
A standard measure of RFID system performance is read-range, which we defined as the maximum horizontal distance from the loop antenna's vertical plane that a tag was detected. The greater the read-range, the greater the total detection field and the less influence that angle and speed of the passing PIT-tag has on the probability of a successful detection. Maximum read-range was assessed by holding a test PIT-tag and slowly moving it through the antenna loop at various points of the configuration. The read-range was measured with a nonmetal measuring tape from the vertical plane of the antenna loop to the maximum perpendicular point that the tag was detected. Read-range was measured after installation and after changes to the system setup or external conditions (e.g., potential noise sources).
As read-range could only be measured in person at the study sites, we had limited capacity to measure the response of readrange to the full variation in noise levels affecting the two systems.
However, the Biomark IS1001 detects and records tag data at a rate of 30 "pings" per second, and so each flight of a tagged bat through the antenna loop is typically logged numerous times. We reasoned that, on average, larger read-ranges (and hence larger detections fields) would result in more logged detections per detection event.
To test our hypothesis that a negative relationship existed between noise and read-range, we calculated the number of consecutive detections recorded for each bat pass and modeled this response variable as a linear function of noise using a zero-truncated poisson regression, implemented with package VGAM with the R software for statistical computing, version 3.5.1 (R Core Team, 2018). All incidents of >50 consecutively logged detections of the same tag number were removed from this analysis, because occasionally very large numbers of consecutive detections were logged (e.g., thousands of detections) likely due to tagged bats roosting near the antenna.

| Detection and encounter probability
We investigated detection probability in situ at Bat Cave with free-flying bats tagged in early February 2017. After tagging, 209 bats were released at night in small batches within the cave, beyond the antenna, between the restriction and the maternity chamber. Each bat therefore needed to fly through the antenna at least once to exit the cave.
Under the assumption that all 209 bats exited the cave after release, detection probability was estimated as the proportion of released individuals detected on the RFID system by midnight following their release, that is, bats were released after midnight, in the early hours of the morning and needed to be detected by midnight on the same date to be included in the proportion detected. This estimate is conditional on noise levels at the time of the experiment and can only serve as a coarse estimate of detection probability (the true detection probability for each animal pass is impossible to quantify directly).
We also used detection histories for each individual to consider how the daily probability of encounter varied with system noise and time of year, using data from the Bat Cave antenna system.
To achieve this, we first derived capture-resight histories for each of the 2,966 PIT-tagged bats, to produce a binary response variable (undetected/detected) for each individual across each day of the study period, with a "day" being defined as the 24 hr between successive middays. Using this variable, we identified the first and last detection event for each individual and derived a second binary variable indicating whether each individual was known to be alive.
We calculated the daily encounter rate as the proportion of individuals known to be alive that were detected each day. We also used a binomial generalized additive model (R package "mgcv") to estimate the per-individual daily probability of encounter (i.e., the probability of being present at Bat Cave and being detected) as a function of noise (averaged for each day) and day of year. For the latter effect, we fitted a cyclic cubic regression spline to ensure continuity of the modeled response between the first and last day of the year.

| System optimization and noise minimization
Noise levels were a major factor in the RFID system performance.
Earthing the system at Bat Cave decreased noise levels, increased read-range (see section 3.2), and resulted in an immediate increase in the number of bats detected per day (Figure 3). Attaching a further two rods in series, on a later date, did not decrease noise levels further. A major source of noise (daily noise ~30%-40%) was inadvertently introduced in September 2016, when park management made changes to a thermal camera in the maternity chamber, approximately 50 m away from the RFID system. The interference caused major disruption to system performance with few bats being detected ( Figure 3). The issue was resolved by disconnecting the camera; thereafter, average daily noise typically ranged between 5% and 18%.
Noise levels recorded by the system installed at Glencoe were lower and less variable than at Bat Cave. The RFID system at Glencoe was initially powered directly from batteries and noise levels averaged 4%. After the solar panel and controller were installed to power the system for long-term use, noise levels became more variable.
There was a daily cycle, whereby average noise only exceeded 5% between dawn and dusk, with a peak at 1 p.m. (Figure 4). The likely source of this noise was the solar controller (which charged the batteries during daylight hours), but this was unlikely to have affected detection success since bat activity at Glencoe is typically recorded between dusk and dawn. Nevertheless, as a precaution, the system was earthed in May 2018 which stabilized the noise levels throughout the day (Figure 4).

| Read-range and the impact of noise
The maximum read-ranges measured at Bat Cave varied under different conditions over the study period and were negatively related to noise ( Figure 5). The highest read-range for this site (89 cm When an earth was installed on the RFID system at the beginning of May, there was a significant increase in the number of bats detected due to lower noise levels. Bats naturally dispersed from Bat Cave soon after earthing and began returning in August. High noise was inadvertently introduced when changes were made to a nearby thermal camera in late September. This dramatically decreased system performance and the number of bats detected. Minor improvements were made when the issue was discovered a week later, including tuning the antenna. The source of the interference was discovered after extensive trouble shooting in late October. The high noise ceased when the camera was unplugged, and the number of bats detected immediately returned to prior levels. A high noise event of unknown origin also occurred on a single date in mid-November

| Detection and encounter probability
The number of individuals tagged in the study and subsequently detected on the system at Bat Cave was lowest in the first year at 92.3%, when noise levels were higher, and improved in the following years when noise levels were lower. Cave after seasonal dispersal periods (Figure 7).

| D ISCUSS I ON
A limitation of PIT-tag technology for wildlife research has been the short read-range capabilities of PIT antennas (Gibbons & Andrews, 2004). With the installation of large RFID antenna systems at southern bent-winged bat roosting caves, we have demonstrated that antenna dimensions and read-range distances can reach greater magnitudes than previously described. Earlier studies using PIT technology at bat roosts or water sources described read-ranges as small as 5-15 cm (Adams & Hayes, 2008;Neubaum et al., 2005), including with the same antenna as used in our study but in a different F I G U R E 4 Mean hourly noise levels (%) at the cave at Glencoe before and after earthing, using all available data (2017-2018)   (Adams & Ammerman, 2015). The greatest read-range we found in the literature for a PIT antenna was 35 cm using a plate antenna (Norquay & Willis, 2014). We have shown that large loop style antenna configurations can achieve read-ranges up to 105 cm on both sides of the detection plane. These results demonstrate greater flexibility of applications for PIT technology to study a wider range of organisms, many of which could not be studied with this technology previously, including many cave-dwelling bat species.
We had high overall detection success, particularly in the second and third years when performance of our RFID system was optimized. Across the full study period, 97% percent of bats were detected at least once. This compares with 76% (Adams, 2015), 67% (Adams & Hayes, 2008) and 62% (Horn, 1998) of bats PITtagged in shorter-term bat studies, and 65% of tagged juveniles and 77% of adult females successfully detected in a longer-term study over four years and multiple roost sites . Factors that may have contributed to the higher overall detection success in our study likely include the advancement in technology used, concerted efforts made to monitor and increase RFID system performance, and the behavior of southern bent-winged bats that show high fidelity to the Bat Cave site and reliably congregate at this maternity cave in large numbers.
Compared with traditional microbat marking and trapping methods, the use of small PIT antennas at roost sites has been demonstrated to significantly increase "recapture" probability and the accuracy of survival estimates, without incurring the cost of increased disturbance from re-trapping . Our data obtained with large antennas F I G U R E 6 Encounter probability models of tagged individuals at Bat Cave in relation to noise levels and day of year. (a) and (b) show the fluctuating noise levels and encounter rates (i.e., proportion of bats detected that are known to be alive) by day of year, with pink bars indicating power outages when no data were recorded on the RFID system. (c) Encounter probability in relation to noise levels (when day of year = 1); (d) Encounter probability throughout the year (using noise levels at 5%). Earthing occurred in early May 2016 Whilst our testing with hand-released bats demonstrated imperfect daily detection rates, mark-recapture methods assume detection/recapture probabilities <1; consequently, this is not usually a problem unless recapture rates are very low (Waller & Svensson, 2016).
A key finding from this study is that large PIT antennas are highly sensitive to noise (unwanted signal) levels. Bat Cave system was notably affected by noise introduced by the power supply. Before earthing the RFID system, unplugging the system's battery charger from mains power increased read-range by an additional 20 cm. This was despite the RFID system running on batteries and the batteryswitcher unit separating the RFID system from the battery charger and associated mains power. Earthing mitigated this issue; however, total collapse in detection capacity resulting from a thermal camera installed 50 m from the RFID system demonstrates the sensitivity of the system to unexpected noise sources, even after earthing. Noise narrow difference between these two measures may have been due to the timing of the release experiment, which occurred at peak season at Bat Cave, when most bats are thought to be present at the maternity cave (Dwyer & Hamilton-Smith, 1965 may be a way around these issues, because noise is a major determinant of read-range ( Figure 5). Given that PIT readers detect tags at a fixed rate per second, the larger the detection field, the faster a tag should be able to pass through the antenna and still be detected. Additionally, greatest read-ranges are achieved with the RFID system when tags pass perpendicular to the detection plane (K. Pomorin, pers. comm.). Ad hoc experimentation upon setting up the antenna confirmed that passing a test tag through the antenna at increasing angles from perpendicular to the antenna plane dramatically decreased the maximum read-range (E. van Harten, pers. obs.).
In fact, holding a test tag parallel to the antenna plane resulted in the tag not being detected at all. Therefore, lowered read-range due to elevated noise would likely compound angle issues, whilst greater read-ranges should allow for a greater range of angles for passing tags. In our study, read-range is likely important for accommodating the natural flight behavior of southern bent-winged bats and may explain the notable differences in encounter probability with small (e.g., 5%) increases in noise levels ( Figure 6c). We have found little literature examining these or other factors affecting PIT-tag detection success (but see Freeland & Fry, 1995 for close-range detection, using hand-held PIT-tag scanners), and we therefore suggest that further investigation into this area is warranted.
Our secondary system at Glencoe was less prone to noise issues than Bat Cave and recorded higher read-ranges throughout the study. However, even under low noise levels (e.g., 5%) at both sites, read-range (and detections per event) was higher at Glencoe ( Figure 5). The higher read-ranges at this study site were therefore not due to average daily noise levels alone. Other factors that may have contributed to the greater read-ranges at Glencoe may be the slightly smaller antenna configuration and that higher noise levels only occurred during the day, when bats were not passing through the antenna to enter or exit the cave (Figure 4). We were initially concerned that metal infrastructure located at the entrance to the cave at Glencoe, including old irrigation pipes and pumping equipment ( Figure 2), might interfere with system performance and successful tag detection. The effect of metal disturbing RFID has been experimentally demonstrated using hand-held PIT-tag readers (Freeland & Fry, 1995) and is highlighted as a potential noise source in the Biomark IS1001 user manual (Biomark Inc, 2015 Overall, compared to alternative methods, PIT-tagging appears to be a safe marking method with favorable benefits to the study population, such as reduced disturbance by minimizing trapping events and low tag weight. Importantly, this technique boasts high re-detection rates and therefore can yield large volumes of continuous data over multiple seasons and years. Whilst the initial cost of equipment may appear as a limitation, this is offset by the comparatively low cost of subsequent re-detections of individuals over the course of a study, especially for larger studies such as ours. One limitation to current PIT-tag studies is the maximum length and potential configurations of commercially available cord antennas. The results of our study, using a 15 m antenna, suggest that even longer antennas may be successfully configured to cover larger entrances. At the time of writing, we have had some preliminary success detecting tagged bats at Bat Cave entrance using a third (specially ordered) 22 m antenna, and as technology progresses options are likely to continue to diversify. Our study demonstrates that large PIT antennas can successfully be used for long-term studies to monitor small, volant, fast-flying animals that move across large distances. The availability of large antennas with larger detection fields increases the potential applications of this technology, and consequently, we believe that the full potential of PIT-tag technology as an ecological research tool is yet to be realized. Engineering.

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to declare.