Emergence and return times in a colonial, cave-dwelling bat: age and sex differences driven by reproductive cycle

The time that bats emerge, and subsequently return, from a colonial roost determines their maximum foraging period and influences their exposure to mortality risks. The order in which different age and sex cohorts emerge and return reflects variation in these cohorts’ resource requirements. The critically endangered Southern Bent-wing Bat (Miniopterus orianae bassanii) is an Australian insectivorous cave-roosting colonial bat. Resource limitation is hypothesised to have contributed to its decline, but may not affect all cohorts equally. We tagged and monitored 3,462 wild Southern Bent-wing Bats over seven years with Passive Integrated Transponder technology. To infer resource requirements of different cohorts over the reproductive cycle, we estimated cohort-specific peak emergence and return times and the frequency of nocturnal returns to the roost. The emergence and return behaviour varied with age, sex, and throughout the annual reproductive cycle. Although adult females and males behaved similarly during the non-breeding period (winter), females emerged significantly (12-21 minutes) earlier and returned (27-62 minutes) later than males during pregnancy, lactation, and weaning. Adult females were less likely than males to be detected overnight in the maternity roost while dependent young were present, suggesting that females prioritised maximising foraging over nocturnal nursing. When juveniles commenced flying, they delayed emergence until several hours after sunset (well after adults had departed the roost). During the 40-day weaning period, they progressively emerged earlier, such that by the end of this period they emerged with the adults, then subsequently foraged for longer than adults over winter. Passive monitoring of emergence and return behaviour in colonial bats can provide valuable data to infer cohort-specific resource requirements. Regular monitoring of a population’s emergence and return times potentially allows for the early detection of changes in resource requirements, and the use of PIT technology allows for the most vulnerable cohort(s) to be identified.

Methods
Study sites

Southern Bent-wing Bats were monitored at a maternity roost and an associated non-maternity roost from 2016 to 2022. The maternity roost is located in Bat Cave, a multi-chambered cave within the Naracoorte Caves National Park World Heritage Area in south-east South Australia. Southern Bent-wing Bats congregate here in the austral spring to form a maternity colony. Occupancy peaks in summer, when the site is occupied by approximately 28,000 adult bats, comprising both females and males of all ages (Hamilton-Smith, 1972). The cave entrance is a collapsed window measuring approximately 8 × 4 m, surrounded by vegetation of varying heights. 

The non-maternity roost is approximately 70 km from the maternity roost. Here, the cave entrance is in a large doline (a funnel-like depression in the ground), and measures approximately 6 × 2 m. The cave is situated on a dairy farm and is surrounded by cleared pastures. Southern Bent-wing Bats are known to move regularly between this site and the maternity roost (van Harten et al., 2022b).

Automated fly-out and fly-in counts at the maternity roost

A monitoring program was established at the maternity roost in late 2021 to estimate the Southern Bent-wing Bat population size each night. An infra-red camera and illuminators are installed outside the cave, which film emergence and return from dusk until dawn. The Southern Bent-wing Bat is the only bat that roosts in this cave. Population estimates are derived using the bespoke N2 software package specifically created to tally emerging and returning bats within each minute of every sampled night at this site. This system provides information about the emergence and return times of the whole aggregation, complementing PIT-tag data, which provides additional information on the behaviour of age and sex cohorts. We primarily used these data to evaluate the accuracy of emergence and return estimates from the PIT antenna positioned inside the cave (see below).

PIT-tagging and detection

A total of 2,966 Southern Bent-wing Bats were trapped and PIT tagged at the maternity cave from 2016-2018, as described in van Harten et al. (2019, 2020). We PIT tagged an additional 496 bats in January and February 2022, following the same method. Bats were trapped throughout the night, from dusk until just before dawn, with two-bank Austbat harp traps (Faunatech/Austbat, Mount Taylor). Bats were sexed based on external genitalia and aged, with juveniles (< 3 months old) classified based on the shape of metacarpal phalangeal joints and presence of cartilage in these joints (Brunet-Rossinni & Wilkinson, 2009). Prior to tagging, bats were scanned for existing PIT tags, then the forearm length and mass of each bat were recorded. Sample sizes were similar for males and females, with marginally more juveniles than adults tagged in later years (795 AF, 671 AM, 969 JF, 1027 JM; Table S1).
In 2022, the PIT tags used were Biomark APT12, with Biomark HPT12 PIT tags used from 2016 to 2018 (van Harten et al., 2019). Both tags are 12 mm glass-encapsulated microchips with a high read range on stationary PIT readers (Karltek 2021, pers. comm.). Tags were inserted subcutaneously to rest between the scapulae, with the insertion site sealed with tissue glue to minimise tag loss (van Harten et al. 2020, 2021). PIT-tag placement was checked before releasing the bats. This technique is associated with high tag retention, with PIT-tag loss of 2.7% in a similar-sized bat, Chalinolobus gouldii, all recorded within 2 months of tagging (van Harten et al., 2021).

A Biomark IS1001 PIT reader system was installed in the maternity cave in January 2016. This system consists of a tag reader, data logger, and a 15 m flexible cord antenna, as described in van Harten et al. (2019). The system was installed in the closest suitably-sized cave passage, which is c. 100 m from the entrance (van Harten et al., 2019) and c. 30 m from the maternity chamber (Dwyer & Hamilton-Smith, 1965). An identical system was installed at the entrance of the non-maternity roost in April 2017 (van Harten et al., 2019). PIT-tag detections are recorded in real-time by these readers, although power outages and high electromagnetic interference have caused temporary loss of data for short periods (van Harten et al., 2019).

All data processing and statistical analyses were performed in R (v. 4.2.2) software for statistical and graphic computing (R Core Team, 2022).

Classification of age-sex cohorts and reproductive periods  

PIT-tagged individuals (detected from 2016 to 2022) were classified into cohorts based on their sex and age at capture. The assumed birth date of first-year individuals was November 24th, the mean date of mass birthing at the maternity roost during the study period (T. Shortt 2023, pers. comm; van Harten et al., 2022b). Juveniles were classified as first-year males or females from their initial capture date until 24/11/[capture year], then reclassified as second-year males or females for the next 12 months. Second-year male and female cohorts were separated from the adult cohorts to examine the effects of age and reproductive maturity separately. After the second year of life, bats were then reclassified as adult males or females on 24/11/[capture year + 1]. Adults and second-year individuals without tags could not be distinguished on capture, so all non-juveniles were excluded from analysis in their capture year and classified as adults on November 24th of the following year. Based on current and past observations of birthing behaviour (Crichton et al., 1989; van Harten et al., 2022b), we defined four key stages of the annual reproductive cycle of Southern Bent-wing Bats: the non-breeding, pregnancy, lactation, and weaning periods (Table 1). 

Emergence and return times

We derived sunrise and sunset times for each monitoring night at each roost using the bioRad package (Dokter et al., 2019). Daytime observations between one hour after sunrise and one hour before sunset were excluded from analysis, as bats detected in this period were presumed to be moving within the cave. 

The emergence and return time of each cohort was estimated separately for each night of PIT-tag monitoring (n=1807 nights at maternity roost, 1256 at non-maternity roost). We defined each cohort’s emergence (and return) time as the peak detection period of individuals in a cohort before midnight, in units of decimal hours since sunset, and return as decimal hours until sunrise. 
To calculate emergence and return times, we tallied the number of unique tagged individuals from each cohort that were detected within each 5-minute interval over the course of each night. To smooth out short-term fluctuations, we then used a moving-average smoother to calculate the mean detection count over every three consecutive (5-minute) intervals. The 5-minute interval corresponding to the maximum moving-average value was deemed the emergence (peak detections before midnight) or return (peak detections after midnight) time for that cohort on that night. 

We analysed emergence and return times at each roost separately, with generalised additive models (GAMs) using the package ‘mgcv’ (Wood, 2011). To compare differences in emergence and return times between cohorts and reproductive periods, we included these fixed factors and their interaction as explanatory variables (Table 2). We also accounted for changes in these times due to the annual progression with a cyclic penalized cubic regression spline on the day of the year and included a random effect of year to consider unexplained variation between years. To account for improved accuracy of nightly peak emergence and return times when more bats were detected, we weighted the GAM model likelihoods for each estimate by the associated moving-average peak count of bats detected (as derived above). Tukey’s tests were used to compare the estimated mean emergence and return times in each period between cohorts at each roost, and between periods for each cohort.

To explore changes in first-year bats’ emergence and return times throughout the weaning period, relative to other cohorts, emergence and return times were also separately estimated for all first-year bats (regardless of sex) and older cohorts throughout the 40-day weaning period at the maternity roost. The emergence and return times of first-year bats and all other cohorts over this period were analysed with separate GAM models that included day of the weaning period and a random effect of year (Table S2). These models were only fitted to data from the maternity roost, as all first-year bats are at this location during the weaning period. 
Emergence times are measured in minutes since sunset, return times are measured in minutes until sunrise. 

Population-level emergence and return times from the PIT and N2 datasets

Peak emergence and return times were also estimated for all tagged bats, regardless of cohort, to estimate the colony-level emergence and return times at the maternity roost on a sample of nights when PIT monitoring and N2 monitoring were both being conducted (199 nights compared for emergence times, 147 nights compared for return times). Peak emergence and return times were then derived from the N2 monitoring data based on detections of all exiting and entering, using the method detailed above. We used paired t-tests to compare in-cave (PIT monitoring) and outside-cave (N2 monitoring) estimates for emergence and return time over this period, and within key reproductive periods (Table 1).  

Overnight activity patterns

To investigate how often bats return to their roosts throughout the night in each reproductive period, we modelled the probability of recording PIT-tagged individuals from each cohort over the course of each night, for each reproductive period (Table 1). First, we calculated the total number of bats recorded on the reader from each cohort on each night (i.e., from an hour before sunset until an hour after sunrise). Second, we tallied the number of bats of each cohort recorded in each five-minute interval over the course of each night. Using the number of individuals of a cohort recorded within each 5-minute interval relative to the number of bats from that cohort recorded that night, we modelled the probability of a bat being detected within each interval over the course of the night using a binomial generalised additive model (GAM) and package mgcv (Wood, 2011). Cohort and reproductive period were included as explanatory fixed factors, along with a random effect of year. To investigate changes in nightly activity patterns between groups over the course of a night, we also included time since sunset as an explanatory variable within this model, using a thin-plate regression spline fitted within each cohort and period combination (Table 2).