The aerial culling trial program P1 occurred from 1–7 in October 2022, covering ~ 20,000 ha of private property in the Limestone Coast region of South Australia, about 300 km southeast of Adelaide (Fig. 1). The program targeted fallow deer – relatively small-bodied cervids with adult masses of 35–55 kg (females) and 50–97 kg (males) (West 2018). For comparison, sambar deer are Australia’s largest deer and weigh around 230 kg (females) and 300 kg (males) (Centre for Invasive Species Solutions 2022a). We reasoned that the small size of fallow deer would increase the likelihood of shotgun pellets effectively penetrating the thorax compared to larger-bodied species.

Figure 1.

Location of the alien deer aerial culling programs in South Australia from 2009 to 2022 (P1–P11). See Table 1 for program descriptions. Red boxes are the minimum convex polygons enclosing all deer kills within each program (P1–P9), or the area searched by helicopters (P10–P11).

All programs used either an AS350 B2 ‘Squirrel’ (Airbus Helicopters, France) or Robinson R44 (Robinson Helicopter Company, U.S.A.) helicopter flown at altitudes generally below 250 m above sea level for all shooting operations. Shotguns can be used up to a maximum of 25 m from the target animal, so helicopters typically remained at 15–20 m above ground level at time of shooting. While rifles have a longer maximum range, shooters in the programs we describe do not typically take rifle shots at distances > 30 m from the target animal.

In P1, one shooter (hereafter, the ‘primary’ shooter) was equipped with a Benelli M2 semi-automatic shotgun with a 26" barrel and a custom choke at full extension, which created a 25-cm pellet spread at 20 m and a 45-cm spread at 30 m. The primary shooter targeted deer in open areas, within a 30-m range. The shotgun was fitted with a red-dot scope (Sightron S30-5 and Aimpoint 9000L); it had a 12-shell tube magazine and was loaded with GB SSG 21-pellet buckshot and Winchester Super-X 16-pellet buckshot. The projectiles of the 21-pellet SSG cartridges have an average weight of 1.8 g, with an average total payload of 37 g. The projectiles in the Winchester Super-X 16-pellet SSG cartridges have an average weight of 2.3 g and a total payload of 36 g. Professional shooters (Wildlife Resources Australia, Wangaratta, Victoria) did not observe any difference in the performance between the different rounds of buckshot. Both round types were mixed into the primary shooter’s ammunition bags, and we did not distinguish between ammunition type during data collection. The primary shooter was positioned in the rear right-hand side of the helicopter behind the pilot (Fig. 2), which gives that shooter the most-efficient position relative to the pilot manipulating the helicopter for optimal distance and angle relative to the target animal.

Figure 2.

Seating configuration of the helicopter crew in P1 A pilot B secondary shooter with rifle and thermal scope C thermographer, and D primary shooter with shotgun and red-dot scope. Yellow and blue polygons show the indicative field of view for the shooters, and the green polygon shows the field of view for the thermographer.

Another shooter (‘secondary’ shooter) was equipped with a Wedgetail WT25 semi-automatic, .308-calibre rifle with a variety of ammunition types. The ammunition included 160-grain copper projectiles used to cull deer near wetlands and creeks. Copper projectiles are being trialled in many pest-control programs in Australia because they do not contain any lead, but they could potentially increase the risk of ricochet (Steven Hess, U.S. Department of Agriculture, Animal and Plant Health Inspection Service, National Wildlife Center, Colorado, personal communication). The secondary shooter targeted deer within vegetated areas and had a range of 70 m. The secondary shooter was positioned next to a thermal camera operator (‘thermographer’; Fig. 2). The thermographer operated a Vayu HD uncooled microbolometer array with the Blackmagic Video Assist and Panasonic GH5 4K video camera and used a high-powered laser to assist the secondary shooter to locate deer in forested areas. The .308-calibre rifle was also equipped with a thermal scope (Pulsar Trail 2 LRF XQ50), so wounded deer in forested areas could be located quickly for follow-up shots and the thermographer could confirm death.

Shooters made chest shots exclusively. For small deer species, especially those that move quickly and erratically such as fallow deer, chest shots are preferred for the best welfare outcomes (Sharp et al. 2022). P1 deployed a deliberate ‘overkill’ policy, which mandated that each deer was shot at least twice (following Hampton et al. 2022). If the target was not moving after a single shot, it would still receive at least one additional chest shot. Two crew members assessed both visually and with the thermal equipment the insensibility/death of each target animal before moving to the next target (see signs for assessing death in ‘Data collection and analyses’). On average, the crew spent 5–10 seconds to determine each apparent death. The total flight time of P1 was 26.3 hours for a total of 611 alien deer culled.

All seating configurations and helicopter operational procedures are obliged to conform to the “Civil Aviation Safety Regulations 1998” and “Manual of Standards” produced and overseen by the Commonwealth of Australia’s Civil Aviation Safety Authority (casa.gov.au). Safety therefore has primacy over all other considerations, including animal humaneness and efficiency/cost components of aerial shooting.

All P1 flights were recorded on the thermal camera and with a GoPro 3 camera. The thermal camera captured all vision from the thermographer’s perspective. The GoPro 3 camera was mounted to the rear firewall of the helicopter and recorded continuously; it captured the activities of all personnel in the helicopter and most of their field of view (Fig. 3). Both systems captured flight audio. The large video and audio files were overwritten every few days, so only a sub-sample of the 611 targeted deer was available for this assessment.

Figure 3.

A GoPro 3 camera, mounted to the rear firewall of the helicopter, captured the seating configuration of the personnel in the helicopter, their field of view, and four deer being pursued (circled in red).

Based on the approach described by Cox et al. (2022), we reviewed all available video footage and audio from the first four hours of flight time on 2, 4, and 5 October 2022 and recorded: (i) number of shotgun and rifle rounds fired; (ii) time taken between the first shot fired at the target with a shotgun and apparent death (with shotgun or rifle); at least two helicopter personnel assessed time of apparent death based on the thermographer observing hotspots indicating that the thorax (heart and/or lungs) had been pierced, and a complete absence of movement determined by any crew member with clear vision; (iii) time between first detection of the target and confirmation of its death; if a deer stayed with its group under pursuit, pursuit time was cumulative for each consecutive deer (i.e., last deer killed in the group was recorded as pursued for the entire time that other deer in the group were being culled); if the group dispersed and a subset of that group had to be re-located, pursuit time was started when the group was relocated.

To test which components of an individual kill explained the most variation in the time from the start of the pursuit to apparent death, we constructed a series of generalised linear models using the glm function in the stats R library (R Core Team 2022). Here, we tested whether the time between first and last/kill shots, number of rounds fired, and group size explained the variation in the time from the start of the pursuit to the kill (with a shotgun). We applied a gamma error distribution and a log link function to account for the non-Gaussian distribution of errors (confirmed appropriate after inspecting quantile-quantile plots), and scaled the response and explanatory variables (except group size) using the scale function in R. We contrasted a total of eight models, including the three additive main effects, all combinations of two additive effects, single effects, and the intercept-only model. We compared the relative probability of the five models per response variable using Akaike’s information criterion corrected for small sample size (AIC_c) (Burnham and Anderson 2002). The bias-corrected relative weight of evidence for each model, given the data and the suite of candidate models considered, was the AIC_c weight (the smaller the weight, the lower the model’s probability) (Burnham and Anderson 2002). We also calculated the percent deviance explained (%DE) as a measure of goodness of fit. We examined model diagnostics using the check_model function in the performance R library (Lüdecke et al. 2021). All data and R code are available at https://github.com/cjabradshaw/deerCullShotgun.

After the morning flights on 4 and 5 October 2022, 20 deer carcasses were located for assessment. Field dissections were done to collect information on shotgun-pellet penetration and spread and organ damage. Shotgun injuries were determined by cutting and peeling back the pelt and visually assessing the external muscle tissue for bruising and penetration of shotgun pellets on the impact and exit sides. Because damage from multiple projectiles to either the heart or lungs is lethal, the number of projectiles that impacted the thorax was also recorded for each carcass.

Following inspection of the muscle tissue and sites of pellet impact, the chest cavity was opened below the sternum using a bone saw. The heart and lungs were removed and inspected for tissue damage, wound channels, bleeding, and blood coagulation to determine whether pellets penetrated the heart and/or the lungs. The heart and lungs were dissected to establish the extent of the wounding by shotgun pellets, if not obvious externally. The chest cavity was also inspected for pooling of blood. All damage was recorded photographically, and the sites assessed for evidence of struggle or distress (such as kicking or disturbance of surrounding ground).

We compared the economic costs and outcomes of P1 to those of 10 other aerial culling programs (P2–P11) completed between June and November 2022. All programs targeted deer in the same region (Limestone Coast) or elsewhere in South Australia, and varied in crew configuration, firearms, equipment, deer density, area covered, and landscape (Table 1). P3, P4 and P5 were part of one large program, but we treated them separately based on their different configurations. We compared the programs according to the following metrics: (i) costs associated with delivering each program, (ii) costs per number of deer culled, and (iii) costs per flight hour and area covered.

Staff costs were included in the assessment because they are necessary to plan and deliver all aerial culling programs. This approach is consistent with ‘competitive neutrality’ requirements for government agencies in South Australia, which ensure government businesses compete fairly in the market (Government of South Australia 2023a). Staff costs were estimated to be $150 per hour for all agencies.

To contextualise any landscape-scale differences among the programs that could have affected cost effectiveness, we also calculated the dominant landcover classes within the area of each program using the South Australia Land Cover raster (2010–2015) at a resolution of 25 m × 25 m (available from data.sa.gov.au/data/dataset/sa-land-cover). We compared the land cover classes in which kills occurred to ‘available’ land cover classes within a minimum convex polygon defined by the locations of all kills in the program. Additionally, we calculated the mean human population density (persons km^-2) within 50 km of the program’s minimum convex polygon to assess the relative likelihood of human visitors to a program area during culls (when near to larger human populations, personnel costs increase – see Results).

We reviewed all available footage from P1, which included 20% of the 611 fallow deer culled (n = 104). Of these, 92% were killed with a shotgun only (n = 96) and 8% with a shotgun-rifle combination (n = 8). Shooters used a total of 383 shotgun rounds and 10 rifle rounds (Table 2).

The mean time between first shot with a shotgun and apparent death was 11.1 seconds (± 0.7; n = 104). Individual deer, or the first deer shot in a group, had the greatest mean time between first shot and apparent death, but this time decreased with subsequent individuals targeted within the group (Fig. 4). The maximum time recorded between first shot and apparent death for any individual deer was 35.9 seconds (Table 2).

Figure 4.

Mean (± standard error) time (seconds) between first shot and apparent death (black circles) and mean (± standard error) pursuit time (seconds) between first detection and apparent death (grey squares) as a function of shot order (either singularly or in groups of 1 to 5).

Mean time between first detection and apparent death was 49.5 seconds (± 3.4; n = 104). Pursuit time increased with subsequent deer shot within a group (Fig. 4). The maximum pursuit time for any deer was 159.0 seconds. See summary data from the analysis of footage in Table 2.

There was a positive effect of deer group size and number of shotgun rounds fired on the total time elapsed since start of pursuit to death (Table 3). These two variables explained ~ 43% of the variation in the response. However, there was no evidence for an effect of the time between the first and last shot and total time elapsed since start of pursuit to death.

The 20 carcasses were recovered and dissected within six hours of being culled in P1. All carcasses had received shotgun wounds only and were located using GPS data collected during the flight. A total of 116 shotgun pellets had penetrated the thorax of the 20 deer (5.8 ± 0.6 pellets deer^-1; range: 3–13 pellets deer^-1). Lethal lung-penetrating wounds were recorded in all 20 animals; 14 (70%) also recorded lethal heart-penetrating wounds. The wounds and their classification are shown in Suppl. material 1. Carcasses showed no indication of struggle or distress or movement from the location at which they were shot and apparent death by the helicopter crew.

For 2022, the cost of delivering 11 aerial culling programs for alien deer in South Australia exceeded $1.1 million (Table 4); the mean ± s.e. cost per program was $100,461 ± $13,385; individual program costs ranged from $45,000 for one component of a larger program (P3) to over $160,000 for P8. As expected, the most expensive component of running any program was associated with helicopter operations, which comprised 54% of all costs.

Operating staff costs accrued by various agencies (South Australian Department of Primary Industries and Regions; Regional Landscape Boards of the Hills and Fleurieu, Limestone Coast, and Eyre Peninsula; National Parks and Wildlife Service; Department for Environment and Water; SA Water; Forestry SA) varied considerably among programs. These costs were largely associated with the location of the operations. P3–P7 occurred on public lands (e.g., parks) near metropolitan areas, so additional staff were required to supervise entrances and prevent public access during the operations. Staff costs for all agencies for all programs combined exceeded $330,000, or 30% of all costs. P6 had the highest staff costs, exceeding $45,000, which comprised 54% of all costs associated with the project. This program required many multi-agency staff to supervise gates and entrances to the operations area, which is a high-profile, peri-urban site on public land (Fig. 1).

From the 11 programs, a total of 3,609 deer (at least 90% fallow deer) were culled during 486 flight hours (see Table 5). In terms of the program cost per deer controlled, P1 was the most cost-effective at $199 deer^-1. The least cost-competitive programs were P10 and P11, which operated in areas with low deer densities (Table 1). Seven animals were culled in P10, costing more than $10,000 deer^-1; P11 cost $27,000 and no animal was destroyed. Excluding P1, the cost per deer controlled in areas with high deer densities (P2–P9) ranged from $210 to $447 deer^-1. The cost per flight hour ranged from around $1,720 (P9) to $8,440 (P7); the mean was $4,526 ± $604 flight hour^-1; P1 cost around $4,950 flight hour^-1. The cost per area covered ranged from around $130 (P9) to $6,800 (P6) km^-2 of program delivered; the mean was $1,445 ± $570 km^-2; P1 cost $868 km^-2.

Deer were most commonly killed in native woody vegetation > 1 m in height (64% of all kill locations across all programs) (Table 5), and in all programs except P7 (Suppl. material 1: fig. S6h), this land cover class was proportionally less-available (20% of area flown) (Suppl. material 1: fig. S6). Sparse native vegetation was the second-most common land cover class in which deer were killed overall (18%), which compares to an availability of only 1% (Suppl. material 1: fig. S6a). Dryland crops was the third-most common land cover class in which deer were killed overall (11%), but this was relatively low compared to an availability of 55% (Suppl. material 1: fig. S6a). Contrary to expectation, there was no apparent relationship between mean human population density within 50 km of a program and either the total personnel costs or personnel costs flight^-1 hour^-1 area^-1 animal^-1; however, the Limestone Coast and Fleurieu Peninsula programs had separate clusters within this cost-population density relationship (Suppl. material 1: fig. S7).

Aerial culling can be an effective, rapid, and humane means for removing large numbers of alien deer (Husheer and Robertson 2005; Bengsen et al. 2022; Pulsford et al. 2023), alien pigs (Cox et al. 2022; Hamnett et al. 2023), and other pest species in vast, remote, and inaccessible landscapes. In 2020, 2021, and 2022, South Australia’s aerial culling programs have removed approximately 3,000 alien deer per year (BDO EconSearch 2022). In addition to aerial culling, some programs have used ground shooting by professional shooters, volunteers and landholders, and commercial harvesting operations (Government of South Australia 2023b). Recreational hunting and culling by private landholders are estimated to remove about 8,300 alien deer annually. With all control approaches combined, approximately 11,300 alien deer are removed per year from South Australia (BDO EconSearch 2022).

Unfortunately, a large proportion of the population of alien deer must be removed each year to drive population decline. For example, at least 34% of the population of fallow deer must be removed each year just to avoid population increase, and even higher culling proportions are required for other deer species (hog: 52%; chital: 49%; rusa: 46%; sambar: 40%) (Hone et al. 2010). The number of fallow deer removed annually from the estimated population of 40,000 in South Australia is around 28% (BDO EconSearch 2022), so the population has continued to grow.

Large-scale, intensive, and coordinated control programs are therefore necessary to drive population declines of alien deer. Improved efficacy of aerial culling programs is clearly needed if management goals to arrest the impacts of deer are to be realised. However, the adoption of new approaches and technologies first requires examination to ensure high animal welfare standards are met, in addition to operational cost effectiveness. Analysis of the outcomes from a recent trial program that used shotguns and thermal equipment, in combination with a rifle, provided insight into the humaneness and effectiveness of a new approach to controlling alien deer in South Australia.

In pest control operations, welfare is generally evaluated in terms of the duration and intensity of suffering (Littin et al. 2004), which inform humaneness assessments of control tools that are common practice in Australia (Sharp and Saunders 2011) and New Zealand (Littin et al. 2004). We used ‘time between first shot with a shotgun and apparent death’ and ‘pursuit time’ as indicators of duration of suffering and penetration and severity of shotgun pellets as indicators of intensity of suffering. The time recorded by Cox et al. (2022) between first shot and apparent death of deer using a rifle was 22 seconds; Hampton et al. (2022) reported that 95% of deer were dead within 57 seconds of the first shot in their program using rifles. In this trial, the average time between first shot with a shotgun and apparent death was 11 seconds, a markedly improved outcome for animal welfare.

Individual deer, or the first deer shot in a group, had the longest mean time between first shot and apparent death, and this interval decreased if targeting subsequent individuals in a group. This decrease is because of the relatively longer time taken to pursue a group of deer after first being sighted, before the first deer is shot. Once the group of deer was engaged, the pursuit time of the remaining deer in the group was usually shorter. The maximum time recorded between first shot and apparent death for any deer was 35.9 seconds, which is an improvement on programs that have used a rifle exclusively (Hampton et al. 2022).

Unlike Cox et al. (2022), our study assessed the metrics of a program that targeted deer with shotguns in relatively open terrain. Shotguns have not been trialled in densely vegetated areas, and so additional trials will be required to determine their efficacy in such habitats. Clearly, different vegetation densities and terrain will affect the outcomes of aerial culling program. The dominant vegetation class of several programs was ‘dry cropland’ (P1–P5, P8–P9), but only P1 also recorded this vegetation type as dominant where deer were killed. Unlike the other programs, outcomes from P1 included a subset of the overall program and selected for shotgun kills, which only occurred in open areas. We found similar proportions of available and kill-location land cover classes in P3–P4 (i.e., including P1, each had 50–60% dry cropland and deer were killed in 30–40% dry cropland; see S1), but the dominant land cover class where deer were killed for most programs was woody native vegetation (i.e., P2–P9) that harbour deer in the landscape.

Other influences such as proficiency of shooters, type of helicopter used, and weather conditions will also affect time between first shot (with shotgun or rifle) and death. In their study, Cox et al. (2022) measured the ‘time from first shot impact to death’, a potentially useful metric for assessing shooter proficiency. We were unable to differentiate impact shots from non-impact shots because the thermographer was not on the same side of the helicopter as the primary shooter with the shotgun. The GoPro footage was not of sufficient quality to assess individual shot impacts. However, we were able to assess overall pursuit time, and time between first shot and apparent death. Cox et al. (2022) and Hampton et al. (2022) recorded pursuit times of around 150 seconds and 90–200 seconds, respectively. The average pursuit time from 104 animals in our study was just 50 seconds, and the maximum pursuit time for any individual was 159 seconds.

In most jurisdictions, procedures and guidelines for aerial culling programs of alien deer dictate that a shot with a rifle is not taken until the shooter has a clear shot of the chest or head, and that there is no risk of a wounded animal escaping to somewhere where a follow-up shot cannot be taken. The spread pattern of the shotgun pellets requires less precision for pellets to hit the thorax of the animal. Hence, using a shotgun reduces the time required to ‘line up’ an accurate and humane shot.

In terms of the intensity of suffering, all animals assessed had received rapid and lethal impacts from shotgun pellets. The average number of thorax-penetrating wounds delivered with the shotgun was higher than in some autopsies of deer culled with a rifle (Hampton et al. 2022). All animals recorded lethal damage to their lungs, and most to their hearts as well. Wounds to the lungs and the pooling and/or clotting of blood in the chest cavity indicated a pneumothorax (collapse of lung) and/or a hemothorax (collapse of lung because of blood in the chest cavity). The wounds to the heart are expected to have caused rapid decrease in blood pressure, rapid loss of consciousness, and rapid death by exsanguination. In combination, these injuries lead to hypovolaemic shock, causing unconsciousness due to inadequate cerebral perfusion pressure, and resultant rapid death from lack of blood supply to the brain (Stokke et al. 2018).

A potential shortcoming of our study is that the apparent death of the target animals in P1 was assessed in the air by the pilot, and at least one other crew member, rather than landing the helicopter to have a veterinary surgeon make a formal assessment (e.g., Hampton et al. 2022). Instead, a veterinary surgeon (A.D.) and a medical doctor (J.D.) were available for consultation for our study. Future research into the use of different firearms to cull deer could benefit from additional veterinary oversight, including work to ensure that culled deer do not have spinal injuries, which could render the animal unresponsive, but alert for some time. In addition, high-resolution photos taken from the helicopter could be used to compare the exact location and position of culled deer with photos subsequently taken from the ground. These records could be used to determine whether there were any signs of movement, distress, or disturbance of the surrounding ground after each deer was killed from the helicopter.

Helicopter-based aerial shooting is a cost-effective tool for alien deer control (Bengsen et al. 2022). However, few studies have assessed the efficiency of different crew and equipment configurations. We assessed a trial program (P1) that used the same pilots, aircraft, and thermal technology as Cox et al. (2022) in their alien pig and deer control research. The main difference was the inclusion of a second shooter armed with a shotgun; it is only the second time (after P5) a program has used a shotgun for targeting alien deer in South Australia.

The largest expense associated with aerial culling is helicopter flight time (Bengsen et al. 2022), largely driven by the cost of aviation fuel. The approximate $2,500 cost hour^-1 of flight time for a B2/B3 Squirrel helicopter is nearly double that of the R44 (approximately $1,000). As such, when using the larger and more expensive helicopters in aerial culling of high-density deer populations, our results indicate that efficiency is maximised by the addition of a thermographer and second shooter with a shotgun. While cost per flight hour and area is relatively high for P1, the efficiency of the configuration was unmatched (25 deer hour^-1 at < $200 deer^-1). Crew configurations would be amended to suit program objectives. For example, a second shooter or thermographer might not be necessary when targeting exclusively open areas where deer densities are high. However, the additional crew members reported other benefits, including (i) additional safety benefits because shooters had opportunities to take brief breaks during each flight; (ii) shooters had the opportunity to change roles when a magazine needed to be changed; (iii) shooters had the opportunity to alternate between using the shotgun and the rifle between flights; (iv) the thermographer had more opportunity to monitor welfare outcomes of targeted animals using the high-resolution thermal camera to confirm death and to locate wounded deer in forested areas; and (v) the thermographer provided a strategic approach to targeting alien deer and enabled searching and scanning areas harbouring deer that might otherwise be missed. The flight crew also reported an increase in the rate of detections of target animals because of the extra spotting capacity from an additional shooter equipped with thermal optics (Rob Matthews, Heli Surveys, Jindabyne, New South Wales, pers. comm.).

Program costs and efficiency will vary with location and density of deer. For example, the cost of targeting sambar deer at low densities in alpine environments exceeded $1,000 deer^-1 (Pulsford et al. 2023). We compared 11 aerial culling programs that varied in location, planning, staffing, and logistic requirements. P10 and P11 occurred in remotes areas with low deer densities. The goal of those programs was to eradicate small satellite populations before they established. The relatively high costs of programs in areas with low deer densities should not discourage land managers, particularly where eradication is possible. Of the programs delivered in areas with high deer densities, program costs ballooned for peri-urban programs because additional staff were required to restrict public access to popular recreation areas. Programs should continue to document the inputs, configurations, and outcomes of their efforts to inform future aerial culling programs of alien deer.