Site fidelity trumps disturbance: aerial shooting does not cause surviving fallow deer (Dama dama) to disperse

Study areas and shooting operations

We captured 71 yearling or adult fallow deer at three sites in New South Wales (NSW), Australia, between June 2020 and April 2022 (Fig. 1). All three sites were characterised by a mixture of woodland and open country used for livestock production on undulating to hilly terrain. However, sites varied in topography, altitude, climate, and proportional woodland cover.

Fig. 1.

Location of three study sites in south-eastern Australia at which 48 GPS-collared fallow deer (Dama dama) were exposed to aerial shooting. There were two aerial shooting operations at A, and one each at B and C.

Deer were captured using helicopter net-gunning and fitted with GPS tracking collars programmed to log hourly location fixes (G5-2D, Advanced Telemetry Systems Inc., Isanti MN, USA; LiteTrack Iridium 420, Lotek Wireless Inc., Newmarket ON, Canada). Capture and handling procedures are detailed in Bengsen et al. (2021) and were approved by the Department of Primary Industries’ Orange Animal Ethics Committee (ORA 20-23-003, ORA 21-24-013). Fallow deer were the dominant deer species at each site, but red, chital and rusa deer were also present at some sites in much lesser numbers. Feral pigs were common at all sites but were less abundant than fallow deer. All sites had been subjected to one aerial shooting operation in the preceding 12 months and all collared deer had been pursued by a helicopter during capture and collaring, so no collared animals were naive to harassment by helicopters. Two does captured prior to Operation A2 were presumed to be a mother-daughter pair due to their relative ages and proximity at time of capture, but there was no indication that any other collared deer was a member of a consistent social group with another collared deer. There was a notable female bias in deer exposed to Operations A2 and C (Table 1) because we avoided capturing deer with large antlers that might be prone to injury while netted (Latham et al. 2020), and because some males in the sample died or lost their collars prior to shooting operations. It was not necessary to avoid large males in the sample exposed to Operation B because they were captured in October 2021 after having shed their antlers. All deer were collared between 2 and 10 months prior to shooting operations.

Deer at each site were subjected to ≥1 aerial shooting operation targeting deer and feral pigs. Seven collared deer that were exposed to Operation A2 also provided data during Operation A1. Based on the age of deer at collaring, we expect that most deer exposed to Operation A2 were present during Operation A1. Shooting operations were conducted by NSW Government shooters in accord with the Feral Animal Aerial Shooting Team manual (FAAST 2020) using Bell 206 Jet Ranger or Eurocopter AS350 Squirrel helicopters, as detailed for NSW sites in previous studies (Hampton et al. 2022; Bengsen et al. 2023). Shooting Operations A1, A2, and B used one helicopter shooting team and lasted for 3 days. Operation C used two helicopter shooting teams operating concurrently for 11 days (Table 1). Shooting teams were instructed to avoid shooting collared deer but could shoot other deer in groups containing collared individuals. Helicopter search patterns generally approximated a Lévy walk, characterised by many small steps within geographic patches and fewer long steps between patches (Supplementary material Fig. S1). Shooting teams searched patches intensively for extended periods and revisited patches throughout the course of the operation. Consequently, we expect that individual deer were repeatedly exposed to disturbance and harassment from shooting teams.

Tracking data

For each aerial shooting operation, we examined approximately 2 months of deer tracking data: 30 days before shooting, the duration of the shoot, and 30 days after shooting. This window allowed us to combine enough data for robust before/after analyses of movement behaviour while minimising the bias due to seasonal variability in deer behaviours. However, Operations B and C were close to the expected peak fallow deer breeding season which usually commences around early April (Mulley 2023). We only considered deer that had tracking data for the complete time window (Table 1). Data were screened to remove implausible location fixes, defined as a movement spike with in and out speed of more than 15 km per hour and a turning angle ≤30°. We also removed any location fixes associated with a horizontal dilution of precision (hdop) >3.

Deer space use

We characterised the spatial activity ranges of deer using biased random bridges (Benhamou 2011) which incorporate the spatio-temporal autocorrelation of an animal’s tracking data in calculating their utilisation distribution (UD). We used the movement-based kernel density estimation on the 99% UD (Benhamou and Cornélis 2010) to describe deer activity ranges before the start and after the end of the aerial shooting operations. We estimated the overlap (between 0 and 1) of the post-shooting range with the pre-shooting range, as well as the relative change in range area (i.e. area post-shoot divided by area pre-shoot) as measures of disturbance from the shooting operations. A relative change in range area with 95% confidence intervals including 1 represented no change, whereas values >1 indicated an increase in range area after shooting and values between 0 and 1 represented a decrease in range area after shooting. For both ranging behaviours (i.e. range overlap and change in area), we fitted generalised linear models (GLMs) with sex, site (K = 3) and their interaction as explanatory variables. Due to the structure of the data, we could not include a year effect in the models. We used a beta distribution to model the range overlap (restricted between 0 and 1) and a Gamma distribution for the relative change in range area (positive values). Deer UD’s were calculated using the ‘adehabitat’ package version 0.4.20 (Calenge 2006) and all GLMs fitted with ‘glmmTMB’ version 1.1.5 (Brooks et al. 2017) in R version 4.2.3 (R Core Team 2023).

To describe the effect of intense disturbance on short-term spatial behaviour, we used an intervention analysis to identify structural changes in mean daily step length (mean hourly distance, MHD; Kay et al. 2017) and the daily maximum distance between any two locations (MxD; Kay et al. 2017) for female and male deer at each site before, during and after being exposed to aerial shooting. For step length, we only used steps with location fixes taken one hour apart. We first estimated the expected MHD and MxD for each combination of site, sex and shooting operation by extracting a time series of movement data starting 30 days prior to the shoot and finishing 30 days after the shoot and using a linear mixed effects model specifying day as a fixed effect and animal ID as a random intercept. Models were fitted using the lme4 package (Bates et al. 2015) in R. For each of the 16 resulting time series, we used the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) unit root test to assess trend stationarity. We then identified breakpoints in the time series at which the linear regression coefficient changed from one stable relationship to a new one using the breakpoints function of the strucchange R-package (v.1.5-2 Zeileis et al. 2002). The optimal number and location of breakpoints for each time series were selected from the model with the minimal Bayesian information criterion. Adequacy of range size GLMs and time series models was assessed using standardised residual plots.

Fallow deer diel cycles

Fallow deer are commonly crepuscular, with peaks of feeding activity around sunset and sunrise (Mulley 2023). This is often associated with deer moving between dense cover that provides shelter and refuge during the day, and more open vegetation that provides richer grazing opportunities at night (Borkowski and Pudełko 2007). All aerial shooting activity occurred during daytime, and deer were often driven short distances out of tree cover into more open areas where they could be more easily shot. We expected deer to adjust the spatial and temporal distribution of their ranging behaviour by: (1) shifting to a more nocturnal activity pattern during shooting; and (2) increasing their use of wooded refuge habitat during daylight hours. For each of the four shooting operations, we used the hourly mean step length to characterise diel activity cycles of deer. We fitted a Gaussian generalised additive mixed model (GAMM) with sex (males and females) and period specific splines (before, during and after aerial shoot) to the step length (m) data. We used a cyclic cubic spline for the hour of the day to account for the circular nature of time and included animal ID as a random effect (intercepts and smooth terms).

To estimate the extent to which deer increased their use of woodlands during daylight hours, we modelled the proportion of tree cover across the three sites. We first used the NSW Native Vegetation Extent 5 m Raster (v1.2, Fisher et al. 2016) to create a 5-m grid with presence-absence (1 or 0) of tree cover. We then resampled the data to a 1-ha grid using pixel-averaging (i.e. sum of all 5-m grid cells/400). For each deer GPS location, we extracted the value of the matching grid cell (0 = no tree cover, 1 = 100% tree cover). For each shooting operation, we fitted a Gamma GAMM with sex and period specific splines (before, during and after aerial shoot) to the tree cover data. True zeros were changed to 0.00001, less than 1/400 = 0.0025 (i.e. the lowest empirical tree cover). We used a cyclic cubic spline for the hour of the day to account for the circular nature of time and included animal ID as a random effect (intercepts and smooth terms). Both diel cycle models were fitted using the ‘mgcv’ package (v1.8.41, Wood 2011) in R and visually checked for an absence of pattern in the model residuals and a k-index close to 1.

Space use

Across the three sites, 48 deer (16 males, 32 females) provided useable location data from 30 days before to 30 days after the respective aerial shoot operations (Table 1). Male activity range areas prior to shooting operations (mean = 777.5 ha, 95% CI: 648.3–906.7 ha) were 2.5 times larger than those of females (mean = 318.9 ha, 95% CI: 227.5–410.2 ha). Expected male activity range area remained stable (i.e. confidence intervals including 1) after three of the four aerial shooting operations and increased 2.7 times after Operation C (Fig. 2). Female activity range area increased 1.4 and 1.7 times after operations A2 and B, respectively, and remained stable after operations A1 and C.

Fig. 2.

Expected mean (and 95% confidence interval) relative change in activity range area (UD 99%) for male and female fallow deer after being exposed to one of four aerial shooting operations. Horizontal dashed line (y = 1) indicates no change, i.e. values above the line show increases in ranging areas and values below the line show decreases in ranging areas after the shoot operations. For operation locations and timings, see Fig. 1 and Table 1, respectively.

Post-shoot activity ranges of all deer overlapped with their pre-shoot range (mean overlap = 79%, range: 12–99%); no deer abandoned their pre-shoot range. The expected degree of range overlap of males varied among shooting operations (Fig. 3), with the lowest overlap after operation C (mean = 41%, 95% CI: 17–66%) and the highest after operation A1 (89%, 95% CI: 80–97%). Activity ranges of females were more stable than those of males, with a mean overlap above 80% for all operations (Fig. 3).

Fig. 3.

Expected mean (and 95% confidence interval) overlap of activity range areas of male and female fallow deer subject to one of four aerial shooting operations. The overlap represents the proportion of the activity range post-shoot that is shared with the activity range pre-shoot. For operation locations and timings, see Fig. 1 and Table 1, respectively.

Step length and maximum daily distance

None of the eight MHD time series showed evidence of a unit root (KPSS test statistic < critical value for α = 0.05). Hence, all these time series were assumed to be stationary and suitable for further analysis. The optimal number of breakpoints for each time series ranged from 0 to 3. Breakpoints coincided with the completion of shooting for female deer in Operation A2 and the onset and completion of shooting for female deer in Operation C (Fig. 4). In Operation A2, expected female MHD increased by 40% from 101.5 m (s.e.m. = 6.2 m) to 142.1 m (s.e.m. = 7.0 m) immediately after the conclusion of the shoot and then returned to initial levels (142.1 m, s.e.m. = 6.0 m) after 8 days. In Operation C, expected female MHD increased by 68% from 89.7 m (s.e.m. = 11.0 m) to 150.5 m (s.e.m. = 11.5 m) with the onset of shooting and then decreased to 101.0 m (s.e.m. = 11.0 m) when shooting ceased. Male deer exposed to Operation C also showed increased MHD during the shoot, but the breakpoints in the time series were offset from the start and conclusion of shooting by 1 and 4 days, respectively. Breakpoints did not coincide with the start or finish of shooting for any other combination of sex and site.

Fig. 4.

Time series of mean hourly distance travelled (MHD) for female (red) and male (blue) fallow deer exposed to one of four aerial shooting operations in eastern Australia. Red and blue ribbons show 95% confidence intervals for each time series. Grey polygons show the timing of shooting operations. Dashed lines represent the fitted values from a model with the optimal number of breakpoints in the time series. For operation locations and timings, see Fig. 1 and Table 1, respectively.

The optimal number of breakpoints for each MxD time series ranged from 0 to 3. For female deer exposed to Operation A2, the expected maximum daily distance between two points increased by 64% from 899 m (s.e.m. = 79.7 m) to 1473 m (s.e.m. = 90.0 m) 1 day after the cessation of the shoot. This increase lasted for 7 days before returning to pre-shoot levels. No breakpoints coincided with the start or completion of shooting for any other combination of sex and site (Fig. 5).

Fig. 5.

Time series of maximum daily distance between any two locations (MxD) for female (red) and male (blue) fallow deer exposed to one of four aerial shooting operations in eastern Australia. Red and blue polygons show 95% confidence intervals for each time series. Grey polygons show the timing of shooting operations. Dashed lines represent the fitted values from a model with the optimal number of breakpoints in the time series. For operation locations and timings, see Fig. 1 and Table 1, respectively.

Diel cycle for step length

As expected, both male and female fallow deer showed strong crepuscular activity (i.e. movements concentrated around sunrise and sunset) before the aerial shoot operations on all three sites (Fig. 6). Deer activity patterns showed site-specific disruptions during the shooting operations. During Operation A1, females showed 60% increase in mean step length during daylight with a shift of movements towards dusk, but their night-time movements remained at the same level. Concomitantly, males abandoned their crepuscular activity for a more even activity distribution across night and day resulting in a 20% reduction in night-time activity (Fig. 6). The only change in activity observed during Operation A2 was a 30% increase in night-time movements for females. During Operation B, females maintained their crepuscular pattern and increased their night-time movements by 30%. During that operation, male movements doubled during the daytime (+115%) and night-time (+110%). During Operation C, females increased their movement by 65% during the day and 40% during the night, and males increased their night-time movements by 20%. After all shooting operations, both male and female fallow deer reverted to their crepuscular activity pattern with similar levels of movements as observed before the shooting operations (Fig. 6).

Fig. 6.

Expected mean (and 95% confidence intervals) diel cycle of movement activity (hourly step length) for female (red) and male (blue) fallow deer exposed to one of four aerial shooting operations in eastern Australia. Vertical dashed lines represent sunrise and sunset. For operation locations and timings, see Fig. 1 and Table 1, respectively.

Diel cycle for tree cover

Our models described clear site- and sex-specific patterns of tree cover use by fallow deer before, during and after the aerial shooting operations (Fig. 7). At site A, before Operations A1 and A2, females used 20–30% denser tree cover than males and both sexes used 20–30% denser tree cover during daylight hours than at night. During Operation A1, while males kept the same pattern, females did not increase their use of tree cover during the day. After the shooting operation, both sexes returned to their pre-shoot patterns of tree cover use. Operation A2 had no effect on female use of tree cover (both during and after). In contrast, males used denser tree cover at night during the shooting operation and reverted to the pre-shoot pattern at the end of the shooting operation but used 10% higher tree cover. At site B, males used tree cover more evenly through the day and night and, although there was no change during the shoot, they used more open areas (10–20% lower tree cover) after the operation, especially at night. At this site, females showed a different pattern, spending the second half of the night in dense tree cover and moving to more open areas (<40% tree cover) during daylight and dusk. During shooting at site B, females kept the same pattern but used 20% denser tree cover at night and returned to pre-shoot levels after the operation. At site C, before the shooting operation, males and females showed similar patterns of tree cover use to deer at site A. During the shoot, males showed a more even use of tree cover at similar levels as before the shoot and kept this pattern after the shoot operation but at a 20% higher level of tree cover. Females at site C maintained their use of tree cover during and after the aerial shoot.

Fig. 7.

Expected mean (and 95% confidence intervals) diel cycle of tree cover use for female (red) and male (blue) fallow deer exposed to one of four aerial shooting operations in eastern Australia. Vertical dashed lines represent sunrise and sunset. For operation locations and timings, see Fig. 1 and Table 1, respectively.

Study limitations

There were several limitations inherent in our sample of collared animals that must be considered when interpreting our results and making recommendations for management. First, all collared deer had previously been exposed to harassment and capture by helicopter, and it is likely that many deer had also been exposed to at least one aerial shooting operation prior to being collared, but our study wasn’t designed to specifically test whether naive deer might have behaved differently. However, as discussed above, we expect that deer that have already been exposed to cues associated with aerial shooting operations should deploy stronger anti-predator behaviour than naive animals. Second, the female sex bias in deer exposed to Operations C and A2 meant that only two and four males, respectively, were exposed to these operations. Future studies using similar methods could reduce sex bias by capturing deer after they have shed antlers and by reducing the interval between capture and exposure to shooting or other treatments. Third, Operation B coincided with the rut, and it is possible that mating behaviour may have obscured some responses to aerial shooting, particularly in males. Fourth, the 60-min interval between location fixes used in our study was too long to identify contacts between individual deer and helicopters, and to characterise immediate avoidance behaviour observed by aerial shooting teams (e.g. Hampton et al. 2022). However, such short-term behaviours are unlikely to cause functionally-important changes in landscape use or disease spread risk unless they accumulate to the point at which they become apparent at coarser time scales that would have been observable in our study. Finally, only deer that were visually assessed to be healthy and in good condition were fitted with collars. Deer that are unwell or in poor condition could behave differently to those studied here, particularly if they are afflicted with a condition that impairs mobility.

Management implications

Contrary to concerns expressed in some wild animal and disease management doctrine (e.g. Animal Health Australia 2023), aerial shooting did not cause any deer to disperse into new areas, and the tendency for deer to become more active at night was not associated with decreased diurnal activity and reduced availability to aerial shooting teams. Inconsistencies in other spatial behaviours between sexes and among operations indicate that some responses of deer to aerial shooting are likely to be context-specific.

Three of the four operations examined were representative of typical practice for population control operations aiming to reduce the impacts of deer on livestock production, using one helicopter shooting team for fewer than 6 consecutive days (e.g. Bengsen et al. 2023). The only clear evidence of behaviour change in these operations was the increased step length in female deer immediately after shooting in one operation, and an increase in activity range area for female deer after two operations. Operation C represented a more intense and sustained disturbance, with two helicopter shooting teams operating concurrently for 11 days. This may be more representative of a disease management operation in which control effort is maintained until a target population density below a disease transmission threshold is reached. Operation C was the only operation in which deer showed a clear and consistent change in spatial behaviour during shooting.

The strong site fidelity evident in collared deer exposed to intense persecution over several days suggests that fallow deer in situations similar to those in our study are unlikely to leave familiar activity range areas if it is not necessary for survival. Consequently, deer that survive initial aerial shooting operations can be expected to remain in situ and be available for follow-up lethal control.