Cost-effectiveness of volunteer and contract ground-based shooting of sambar deer in Australia

Keywords: biological invasions, catch per unit effort, Cervus unicolor, contract shooters, CPUE, culling, GPS, invasive species, ungulates, volunteer shooters, wildlife management.

Overabundant deer can have significant undesirable impacts on native biodiversity, primary industries and human health (Côté et al. 2004; Martin et al. 2020; Spake et al. 2020; Carpio et al. 2021), including in Australia (Moriarty 2004; Davis et al. 2016; Cripps et al. 2019; Davies et al. 2020a). Mitigating the undesirable impacts of deer often involves reducing their density using ground-based shooting (Caughley 1983; Frost et al. 1997; Bennett et al. 2015; Bengsen et al. 2020). Ground-based shooting for population control (as distinct from harvesting) can be undertaken by volunteers (i.e. unpaid) or professionals (i.e. contractors or government/agency employees; Burt et al. 2011; Warren 2011; Bengsen et al. 2020). Volunteer shooters give their time and experience without financial return, although non-monetary compensations (e.g. accommodation, fuel and food) are sometimes provided to them. Volunteers are generally local recreational hunters motivated by the experience of hunting in areas that might otherwise be unavailable (Gidlow et al. 2009), the opportunity to collect meat (Grilliot and Armstrong 2005), social interactions (Black et al. 2018), or by the satisfaction of being part of management programs (e.g. pest management, conservation; Woods et al. 1996; Finch et al. 2014). In addition to those motivations, contract shooters are motivated by financial incentives (usually a fee per unit of time). Contract shooters generally have more experience in control operations and already possess the licences and insurances required to participate in management programs. A recent systematic review reported that only 30% of mammalian control operations that relied only on volunteer shooters or recreational hunters achieved their management objectives (Bengsen et al. 2020). However, the effectiveness of different types of ground shooter (i.e. the number of animals killed per unit of time) have not been compared within a management program in Australia.

The equipment used for ground-based shooting can improve efficiency of operations through increased shooting opportunities and reduced shooting distances. Since deer are often most active when it is dark (Bentley 1998; Leslie 2011), white-light spotlights (Hampton et al. 2022) and thermal-vision equipment (Logan et al. 2019) can make ground-based shooters more effective. Sound suppressors reduce the peak noise level of a gunshot away from the line of fire, potentially reducing the fleeing behaviour of peripheral animals and hence providing the shooter with additional shooting opportunities (Williams et al. 2018). Sound suppressors also reduce the peak pressure at the shooter’s ear, providing safer conditions for long and repetitive shooting operations (Murphy et al. 2018). In dense vegetation, an indicator dog can greatly increase the success of deer hunters (Novak et al. 1991; Godwin et al. 2013). A dog can also reduce the time to find deer that have been shot (i.e. to confirm the death of the animal), further increasing efficiency.

The efficiency of ground-based shooting operations can be measured as the number of animals killed per unit of time or catch per unit effort (CPUE), a widely used index to monitor deer population dynamics (Batcheler and Logan 1963; Haskell 2011; Gürtler et al. 2018). When deer densities increase, CPUE will increase until the system eventually reaches saturation, i.e. the effort required to process each kill limits the number of deer shot per unit of time. Conversely, when deer density decreases, CPUE will decrease because it becomes more difficult for the shooter to find a deer (Van Deelen and Etter 2003). This relationship between density and CPUE depends on the landscape and the environment (Fraser and Sweetapple 1992; Nugent and Choquenot 2004). Increasing road density was associated with increasing probability of success for white-tailed deer (Odocoileus virginianus) hunters in the United States (Lebel et al. 2012), and low temperatures and rain limited red deer (Cervus elaphus) hunting effort in Norway (Rivrud et al. 2014). Most studies have focused on recreational hunting, and little information is available for contract and volunteer shooters involved in management of invasive or overabundant deer.

Managers seeking to manage deer populations by ground shooting want to know how the costs of removing deer differ between contract and volunteer shooters (i.e. the cost per unit outcome). If effort and cost are standardised, then CPUE can be transformed into the cost per unit outcome (i.e. the cost of removing one individual from the population; sensu Cook et al. 2017), which can then be compared between shooter types. The costs of management programs using ground-based shooting should include the administrative and organisational costs, the operating costs (i.e. supervision, financial and in-kind compensations), and the equipment purchased for the program (e.g. ammunition, vehicle, night-vision equipment). Reporting of transparent and standardised cost per unit effort have been limited for ground-based shooting of deer. DeNicola et al. (2000) reported costs of US$91–310 per white-tailed deer across several management programs for white-tailed deer by using professional shooters in urban areas in the United States, but did not detail how those values were calculated. Hubbard and Nielsen (2011) used a transparent costing tool to show that the cost per deer for volunteer shooters (peri-urban white-tailed deer) varied from US$37 to US$122 depending on the management strategy used. In Australia, the standardised costs and effectiveness of ground-based shooting have not been documented for any of the six deer species present.

Sambar deer (Cervus unicolor), the largest species of deer in Australia (220 kg for males, 140 kg for females), have colonised a wide variety of landscapes since they were first released in 1860 (Bentley 1998; Moriarty 2004). In Victoria, sambar deer are listed as a key threatening process for the biodiversity of native vegetation (Victorian Flora and Fauna Guarantee Act 1988) because of herbivory, antler rubbing, trampling and wallowing. There is a long history of recreational ground-based shooting of sambar deer in Victoria (Bentley 1998; Harrison 2010). For example, during our 5-year study period (see below), an average of 88 849 sambar deer were reported killed annually across the state by licenced hunters (range: 55 094–131 258; Moloney and Flesch 2021). Despite these sustained, large recreational hunter harvests, the Victorian sambar deer population is increasing (Forsyth et al. 2018; Watter et al. 2020). Professional ground-based shooters are increasingly used to reduce the abundance of deer in Australia (e.g. sambar deer management in one of Melbourne’s water catchments; Bennett et al. 2015), but the costs of such programs are not reported. There is substantial interest in using recreational hunters as volunteer shooters in organised Victorian deer control programs, from both hunters and public land management agencies, partly because of perceived cost-effectiveness benefits over professional shooters. However, it is recognised that co-ordinating volunteer shooters has a cost for the co-ordinating agency (New South Wales Natural Resources Commission 2017; Parliament of Victoria Environment, Natural Resources and Regional Development Committee 2017).

In this study, we compare the cost-effectiveness of volunteer and contract ground-based shooters in a 5-year management program that aimed to reduce the impacts of sambar deer in Alpine National Park, Victoria. On the basis of prior knowledge from the literature, we tested the following four hypotheses: (1) that the higher kills per unit time of contract shooters than of volunteer shooters outweighs their higher cost-per-unit-time; (2) that contract shooters cover larger areas per unit time than do volunteer shooters, resulting in more shooting opportunities; (3) that supporting equipment, particularly a vehicle with a spotlight or thermal vision when walking, improves the effectiveness of volunteer ground-based shooting; and (4) that ground-based shooting is most cost-effective around dusk, when sambar deer are most active in this environment (Comte et al. 2022).

The Alpine National Park deer control trial was implemented in two blocks in Alpine National Park, the Bogong High Plains (BHP) and the Howitt-Wellington Plains (HWP). Each block consisted of four management units (Fig. 1) randomly assigned to treatment (organised ground-based shooting) or non-treatment (no organised ground-based shooting). The proposed design included the deployment of motion-sensitive cameras in each management unit for the duration of the program (Davis et al. 2015); however, owing to financial constraints, only two units in BHP were monitored, one treatment and one non-treatment (Fig. 1). The two blocks were selected to be spatially independent (>50 km apart) but similar in terms of landscape and habitat. Both blocks were especially chosen for the presence of alpine sphagnum bogs and associated fens dominated by Sphagnum subsecundum and S. cristatum (hereafter referred to as alpine peatlands), a nationally endangered community notably threatened by sambar deer (Australian Department of the Environment 2015). Alpine peatlands were generally present on flat areas surrounded by alpine and subalpine woodlands dominated by snow gums (Eucalyptus pauciflora). Montane wet forests (dominated by alpine ash E. delegatensis) were the dominant community in steeper slopes near watercourses, usually with a dense understorey (Conn 1993).

Fig. 1.  Map of the two blocks with treatment (organised ground-based shooting) and non-treatment (no organised ground-based shooting) management units in Alpine National Park, Victoria, Australia, 2015–2020.

The BHP block surrounded the Falls Creek ski resort (36.87°S, 147.28°E). The four management units within this block were 3956–4422 ha in area, with the two treatment units separated by >7 km (Fig. 1). The road network in the two treatment units consisted of unsealed fire trails and management vehicle-only tracks (i.e. four-wheel-drive tracks closed to the public by locked gates). All roads were open to mountain biking and hiking, and there were several campgrounds and walking tracks in each treatment unit. Elevations within the four units ranged from 720 to 1882 m asl. The mean annual rainfall was 1302 mm, with monthly maximum and minimum temperatures ranging from 8.3°C and 16.9°C in February to −3.2°C and 0.9°C in July (Australian Bureau of Meteorology Falls Creek weather station, 1765 m asl). Snow was present between June and October, with a maximum mean monthly depth of 167 cm (Daily Snow Depth Records Falls Creek, Department of Environment, Land, Water and Planning, https://discover.data.vic.gov.au). Recreational hunting was prohibited in this block.

The HWP block (37.33°S, 146.77°E) was ~60 km south-west of the BHP block (Fig. 1). The four management units ranged from 3842 to 4681 ha, with the two treatment units being separated by >6 km (Fig. 1). The road network in the two treatment units consisted of unsealed two-wheel drive roads and four-wheel-drive tracks that were accessible to the public, and several management vehicles-only tracks closed to the public. There were several campgrounds in each treatment unit. Elevations within the four units ranged from 524 to 1709 m asl. The mean annual rainfall was 1622 mm and the minimum and maximum daily temperatures ranged from −7.4°C to 0.4°C and from 6.1°C to 27.0°C in August and January, respectively (Australian Bureau of Meteorology Mount Buller weather station, 1707 m asl.). There were no snow-depth data for this block. In contrast to the BHP block, recreational hunting of sambar deer was authorised annually from 15 February to 15 December, and was restricted to stalking on foot during the day without a dog (Victoria Game Management Authority, https://www.gma.vic.gov.au/hunting/deer).

This paper focuses on the ground-based shooting operations conducted in the four treatment units (two in each block). We have previously shown, using images of sambar deer activity from motion-sensitive cameras deployed in BHP, that during this management program the modal sambar deer group size was one and that a maximum of eight deer were detected together (Comte et al. 2022). Sambar deer detections were higher in dense woody vegetation and nearer to roads and tracks, which, owing to low vehicle traffic, may act as movement corridors. The diel activity pattern of sambar deer was mostly crepuscular (higher activity at dusk and dawn) with a larger peak of activity at sunset. The annual accumulation of snow between July and September was associated with a strong decrease in detections as sambar deer moved to lower elevations (Comte et al. 2022).

During the management program, all ground-based shooting was restricted to time-limited operations (n = 32 operations in BHP, n = 9 in HWP). Because both blocks remained open to outdoor activities (including hiking, horse riding and cross-country skiing) during the shooting operations, potential visitors were advised of scheduled shooting operations by signs erected at key access points. Each operation was divided into shifts (median = 8; range: 1–16), defined by a start and end time (median = 5.5 h; range: 0.8–12.0 h), during which teams of shooters (up to four people) were given access to one management unit (Table 1). For each shift, the teams of shooters recorded the number of deer seen and killed by sex–age classes defined as adult male, adult female, yearling male, yearling female, or calf (no sex assigned). Shooters used a hand-held GPS (Garmin GPSMAP 78s, Garmin Ltd, Olathe, KS, USA) to record their movement track-log during the shift and the location and time that each deer was killed.

Table 1.  Ground-based shooting operations for sambar deer management in Alpine National Park, Victoria, Australia.

The 75 volunteer shooters were members of the Australian Deer Association and/or the Sporting Shooters’ Association of Australia. Each volunteer shooter possessed valid firearm and game licences for the state of Victoria as well as an authorisation to possess, carry and use a firearm in a National Park (Section 37 of the National Parks Act 1975 Victoria). Volunteer shooters were accredited by their respective organisation to participate in deer management programs after the completion of a theoretical and practical course. In Victoria, sambar deer hunters are legally required [Wildlife (Game) Regulations 2012] to use a minimum calibre of 0.270 (6.85 mm) with a minimum projectile weight of 130 grains (8.45 g). Eligible volunteer shooters were recruited for each operation on a first-come, first-served basis. For all operations, volunteer shooters used their own hunting equipment including rifle, ammunition, four-wheel-drive vehicles and spotlights. Shooting teams were provided with a hand-held GPS unit (as above) and Victorian StateNet Mobile Radio Network radios. Volunteer shooters had access to two thermal hand-held monoculars (Helion XQ, Yukon Advanced Optics Ltd, Lithuania) and two detachable thermal scopes (Pulsar Trail XP50 LRF, Yukon Advanced Optics Ltd, Lithuania).

In the first year of the program (May 2015–May 2016), the contract shooters were selected by a market-based expression of interest and merit-based procurement process. One company was selected (Strathbogie Wildlife Pty Ltd, Holbrook, New South Wales), and their contract was renewed for subsequent years. There were six contract shooters, all with considerable experience in ground-based shooting of sambar deer in south-eastern Australia. They possessed a valid corporate firearm licence for Victoria, including public liability insurance (Firearms Act 1996 Victoria). The contract shooters used the following three calibres: 0.338 Winchester Magnum, 0.308 Winchester and 0.300 Winchester Magnum. All contract shooters held an authorisation to possess, carry and use a firearm in a National Park (Section 37 of the National Parks Act 1975 Victoria) and a permit for using sound suppressors (Section 57 of the Firearms Act 1996 Victoria). For all their operations, contract shooters used their own hand-held GPS, hand-held thermal monoculars and thermal scopes (makes and models similar to those provided to the volunteer shooters). Contract shooters occasionally used vehicles (four-wheel drive or all-terrain vehicle) or horses to move within management units. Two of the contract shooters used gundogs to find shot deer.

Following Iacona et al. (2018) and Bengsen et al. (2022), we grouped costs on the basis of their scaling potential (Supplementary Table S1). The ‘organisational costs’ represent fixed annual labour time and therefore scale down with increased ground-shooting effort during the year. The ‘operational costs’ were calculated as the labour time and financial compensations for each ground-based shooting operation. These costs varied between volunteer and contract shooters and between blocks. Operational costs scale up with increased effort because they are proportional to the number of management operations. Finally, the ‘equipment costs’ summarise the purchase of assets necessary for the ground-based shooting operations. These one-off costs scale down with the amount of effort. All costs are reported in 2021 Australian dollars (A$). The labour time included different staff salary levels plus on-costs, so we used an average hourly rate of A$57.0 (Parks Victoria Enterprise Agreement 2016, https://www.fwc.gov.au/documents/documents/agreements/fwa/ae423611.pdf).

During the first year of the program, all shooting operations (75 shifts within eight operations) were conducted in BHP Unit B by volunteer shooters only. Only one operation was performed by contract shooters, in BHP Unit D in May 2016, and this operation was excluded from our analysis. During this first year, volunteer shooters used different methods for ground-based shooting. The majority of shifts consisted of stalking on foot during daytime (n = 51), of which 12 used a gundog to indicate sambar deer (Table 2). An additional 21 shifts were conducted at night, shooting from a vehicle with a spotlight (n = 13) or stalking on foot using spotlight or thermal-vision equipment (n = 3 and n = 7 respectively). Two shifts consisted of static glassing (i.e. long-distance shooting during the day from one location) and one shift consisted of a drive (i.e. people flushing deer toward waiting shooters) involving 12 shooters. These three shifts were removed from analyses because of insufficient replication. We compared the CPUE (i.e. the number of deer killed per hour of shooting operation) of each method by using a Bayesian negative binomial generalised linear mixed model (GLMM). We used the number of deer killed (i.e. the catch) as the response variable C_i, with p_i the success parameter and r the dispersion parameter:

Table 2.  Ground-based shooting methods and equipment used by volunteer shooters in BHP Unit B in Alpine National Park, Victoria, Australia, during May 2015–May 2016.

We included the effort (i.e. duration of the shift in hours) as log (duration_i), an offset (i.e. parameter fixed at 1.0) such that λ_i was the expected (mean) CPUE. The linear predictors included five shooting methods (Table 2, excluding glassing and day drive) as dummy variables (0 or 1), with spotlighting from a vehicle as the reference level (i.e. the intercept α). We included the size of the team and the duration of the shift (hours) as linear predictors and θ_{operation i} was the random effect of the operation (n = 8).

In subsequent years (December 2016–February 2020), the ground-based shooting operations on both blocks were organised annually between November and May (spring–autumn), with volunteer and contract shooters alternately accessing the treatment units (with the exception of BHP Unit D, which was restricted to contract shooters). Ground-based shooting ceased in BHP in February 2019, and in HWP in February 2020. On the basis of the experience gained during the first year of the program, all ground-based shooting operations were conducted during night-time by stalking on foot using thermal-vision equipment or from a vehicle with spotlight or thermal vision equipment. We used a negative binomial GLMM, as previously described, to model the temporal trends in number of deer killed (the catch) with an offset of the duration of the shift (the effort). The model was fitted with different intercepts for contract and volunteer shooters, and with different slopes for those two shooter types for each block (n = 4, i.e. contractors and volunteers on BHP and HWP, respectively) over time (days since first shooting operation). We added the cumulative number of deer killed during each annual shooting period (November–May), and the number of deer seen during each shift as linear predictors. The shooting operations were undertaken on four different units; we therefore used a nested random effect of the operations (n = 35) within units (n = 4).

We implemented both models in a Bayesian framework by using gamma priors for the dispersion parameters r_m ~ Gamma (0.01,0.01) and diffuse normal priors α_m, β_m ~ N(0,10) for intercepts and linear predictors, respectively. We modelled the random effects by using normal priors for the group-level mean θ_n ~ N(0,σ_n) and a uniform distribution for the group-level variance σ_n ~ U(0,20). The models were fit using the packages runjags ver. 2.0.4-6 (Denwood 2016) and coda ver. 0.19-3 (Plummer et al. 2006). We ran four Markov-chain Monte Carlo (MCMC) chains with a thinning of 10 for a total sample size of 100 000, with 5000 adaptation runs and 5000 burn-in runs. We assessed the mixing of the MCMC chains visually and with the Gelman–Rubin diagnostic (; Gelman and Rubin 1992).

We visually inspected all track-logs and discarded any track-log with GPS malfunction or loss of satellite, that is, inconsistencies in the frequency of fixes or multiple gaps in the track log larger than 30 min. Because each hand-held GPS unit recorded the location of the shooter at different frequencies (<5 to >120 s), we standardised the tracking data by discretising all track-logs with a 5-min step duration. For each team of shooters, we calculated the total distance (m) covered during the shift (sum of all step lengths) and the mean speed (metres per second) during the shift by using ArcGIS (ver. 10.6.1.9270, Esri, Redlands, CA, USA; www.esri.com) and the R-package adehabitatLT ver. 0.3.25 (Calenge 2006).

A high frequency of fixes can produce serial autocorrelation of the relocations in tracking data (Boyce et al. 2010). We therefore used the biased random bridge (BRB; Benhamou and Cornélis 2010; Benhamou 2011) approach to estimate the shooters’ utilisation distribution (UD). Instead of considering each new step in a trajectory as a random decision (purely diffusive movement) as in the Kernel Brownian bridge estimators (Horne et al. 2007), the BRB incorporates a directional drift parameter (advection) between relocations. This approach is therefore more realistic when modelling movements of animals, including people (Papworth et al. 2012). We pooled the UDs for volunteers and contract shooters separately on each of the four treatment units using a 1-ha grid. Because of the difference in equipment and strategy (i.e. time of day and duration of the shifts), we discarded the UDs from the first year of the program on BHP Units B and D. As commonly used for mammal space use (Pellerin et al. 2008), we considered the core area (UD50, the smallest area containing 50% of the space-use probability) and the main space use during the study (UD95). We also defined an extended space use (UD99) including the least-used areas (Horne et al. 2007). Any cell not included in one of the categories was considered as not used by shooters during the study. We measured the overlap between the space use of volunteer and contract shooters for each level of UD on each management unit (BHP-B, HWP-B and HWP-D) as the number of grid cells (ha) common to both shooter types’ UDs. All space use calculations were performed with the R-package adehabitatHR ver. 0.4.18 (Calenge 2006). We used the UDs to fit separate resource utilisation functions (RUF; Rowland et al. 2021) for volunteer and contract shooters. We restricted the RUF to BHP-B because it was the only unit on which sambar deer activity was simultaneously monitored using camera traps (Comte et al. 2022) and was also the unit with the highest combined effort for volunteer and contract shooters (Table 1), therefore remaining representative of each shooter type. To reduce the spatial autocorrelation of UD between proximal grid cells, we randomly selected two sets of 100 grid cells (one set for volunteers and one set for contractors) across the management unit. We created our response variable as the UD value (i.e. the UD probability value) extracted from the randomly sampled grid cells. For each sampled grid cell, we calculated the distance (m) from its centre to the nearest road and to the nearest watercourse, the mean slope (decimal degrees) on the grid cell and the mean probability of the grid cell being covered by woody vegetation (scaled as 0–100). We also calculated the distance (m) to the nearest alpine peatland as the main ecological asset targeted by the ground-based shooting. Last, we included the predicted sambar deer activity previously modelled on camera-trap detections (i.e. predicted monthly detections of sambar deer in the absence of snow cover; Fig. 2c). Detailed descriptions of the environmental covariates and their sources are given in Comte et al. (2022).

Fig. 2.  Space use of (a) volunteer and (b) contract deer shooters (79 and 32 shifts between 2016 and 2019, respectively), and (c) predicted sambar deer activity (monthly detections) based on camera-trap monitoring, 2015–2019, in unit BHP-B, Alpine National Park, Victoria, Australia. Shooters’ utilisation distributions (UD) were calculated using the biased random bridge method (BRB) with GPS locations collected every 5 min. See Supplementary Fig. S1 for the shooters’ space use in the other treatment units.

The UD probabilities in our response variable included true zeros. We therefore fitted a zero-inflated beta regression (Ospina and Ferrari 2012) to the UD probabilities y_i, as follows:

where p_i is the probability of y_i = 0; μ_i and ϑ_i are the mean and dispersion parameter of the beta distribution when y_i ∈ (0,1). We assumed a constant dispersion parameter for the beta distribution log(ϑ_i) = c, such that:

with β_mX_mi the linear predictors. The mean UD on grid cell i was therefore:

As we used all six explanatory variables described previously for both the mean and the zero probability of the beta distribution, X₁ = X₂ in our model. We fitted the model in a Bayesian framework by using diffuse normal priors β_m ~ N(0,10) for all model parameters. We ran three MCMC chains of 50 000 samples each, with 1000 burn-in runs. We assessed the mixing of the MCMC chains visually and with the Gelman–Rubin diagnostic (). The regression analyses were performed using the packages zoib ver. 1.5.5 (Liu and Kong 2015), runjags ver. 2.0.4-6 and coda ver. 0.19-3.

Deer are potentially more easily detected by shooters when they are more active (Little et al. 2014; Rivrud et al. 2014). Therefore, we expected a higher CPUE at dusk and dawn, when sambar deer were most active in our study area (Comte et al. 2022). We calculated the daily distribution of effort for both volunteer and contract shooters as the number of shifts active during each hour of the day across all treatment units and blocks. We then calculated the CPUE for each type of shooter as the total number of deer killed each hour divided by the total effort. We considered CPUE only for hours with at least five shifts. We scaled the times to sunrise (06:00) and sunset (18:00) by using the function sunTime in the R-package overlap ver. 0.3.3 (Ridout and Linkie 2009).

The Alpine National Park deer control trial was conducted by Parks Victoria as part of a pest management program under Section 28A of the Wildlife Act 1975 (Victoria), and therefore did not require additional Animal Ethics Approval for the ground-shooting operations. For the Bogong High Plains, the Department of Environment, Land, Water and Planning granted Parks Victoria annual ‘Authorisation to destroy, possess, and dispose of protected wildlife and to use a gun to gain possession of protected wildlife’ (reference numbers: 14433193, 14484589, 14567926, 14637928, 14709708, 14733000). For the Howitt-Wellington Plains, the Game Management Authority issued Parks Victoria with an ‘Authority to control or destroy game’ (reference numbers: GMA/2017/07, GMA/2017/01). The use of motion-sensitive cameras was approved by the Northern Melbourne Institute of TAFE Animal Ethics Committee (approval numbers: 01/14 and 03/17).

Between May 2015 and February 2020, 573 sambar deer, in total, were killed during 2135 h of ground-based shooting in the four management units (total area: 17 412 ha), 21% (during 1051 h) by the volunteer shooters and 79% (during 1084 h) by the contract shooters (Table 1). More male (61%; 49% adults and 12% yearlings) than female (39%; 35% adults and 4% yearlings) sambar deer were killed, but the sex–age classes of the sambar deer killed were similar for both shooter types (Chi-squared test: χ² = 4.11, d.f. = 3, P = 0.25). Seven calves (sexes not recorded) were killed, and sex–age class information was not recorded for 56 other deer (50 killed by contract shooters and six by volunteer shooters). On average, contract shooters moved faster (n = 143 shifts with complete GPS track logs, mean = 0.76 m/s, 95% CI: 0.67–0.86) than did volunteer shooters (n = 140 shifts, mean = 0.41 m/s, 95% CI: 0.35–0.46) during the shooting operations.

During the first year of the program, 26 sambar deer were shot by volunteer shooters in unit BHP-B. Most sambar deer were killed at night from a vehicle by using a white-light spotlight (n = 16) or on foot using thermal-vision equipment (n = 6). Three deer were killed during daylight stalking, all while using gundogs. The last deer was killed during the 1-day drive that involved 12 volunteer shooters for 1.82 h, and this datum was not included in our analysis. Considering a mean effort (duration = 3.72 h and shooting team size = 2), the CPUE for ground-based shooting of sambar deer was highest at night when using a spotlight from a vehicle or stalking with thermal-vision equipment, both being four times more effective than day stalking with gundogs (Table 2). All three methods were significantly more effective than day stalking without gundogs. No deer were killed by night stalking (on foot) using a white-light spotlight, but because of the small sample size, the credible intervals were large (Table 2). Each additional hour of ground-based shooting resulted in a 56% (95% credible interval (CrI): 7–147) increase in CPUE, and the size of the shooting team did not have a significant effect on CPUE (Table S2).

During the following years (December 2016–February 2020), while using similar methods (equipment and schedule), the CPUE was 4.58 (95% CrI: 2.07–8.64) times higher for contract shooters (0.77 deer per hour, 95% CrI: 0.21–2.42) than for volunteer shooters (0.18 deer per hour, 95% CrI: 0.10–0.28; Table S3). During that time, there was no evidence of a change in CPUE for the contract shooters (credible intervals included zero for both BHP and HWP), but volunteer shooters became more effective over time in HWP (annual CPUE increase of 55%; 95% CrI: 15–97%; Table S3), a trend not observed in BHP. For both shooter types, the CPUE increased by 20% (95% CrI: 16–26%) with each additional deer seen but was not affected by the cumulative number of sambar deer removed during each annual shooting period.

Of the 309 shifts between November 2016 and February 2020, 236 (76%) provided GPS track-log data suitable for analysis (75% for volunteer shooters and 77% for contract shooters). Both volunteer and contract shooters exhibited highly heterogenous space use in the four treatment units (Figs 2a, b, S1a, b), with core areas (UD50) corresponding to 14% (range 10–17%) of the total area used (UD99). In all four treatment units, the area used by contract shooters was greater than that used by volunteer shooters. There was 80% (range 66–92%) overlap between contract and volunteer shooters space use for the UD99, but only 40% (range 26–58%) for the UD50 (Table 3).

Table 3.  Pooled utilisation distribution (UD) of volunteer and contract deer shooters, and their overlap, in four management units in Alpine National Park, Victoria, Australia, 2016–2020.

The RUF for ground-based shooting in the treatment unit BHP-B showed that the space use of both volunteers and contractors was significantly influenced by the landscape. The zero-inflation component of the RUF (Fig. 3a, Table S4) showed that both shooter types were more likely to use areas close to alpine peatlands and roads. Volunteer shooters also preferentially used flatter areas, whereas contract shooters used areas closer to watercourses. In addition, the beta distribution component of the RUF (Fig. 3b, Table S4) showed that within the areas of the treatment units that were used, volunteers used flatter areas and areas closer to roads and watercourses more intensively than they did steeper slopes and areas further away from roads and watercourses. However, contract shooters showed no significant associations with spatial covariates within the areas that they used.

Fig 3.  Posterior distributions of the parameters of the (a) zero-inflated and (b) beta distribution model for the resource utilisation function (RUF) for space use of volunteer (blue) and contract (red) shooters in Alpine National Park, Victoria, Australia, 2016–2020. For clarity of the figure, posterior distributions of distance parameters are expressed for 100 m. See Table S4 for full model outputs. Circles and triangles are means, bars are 95% CrIs.

The time of shooting was recorded for 451 (79%) of the 573 deer shot during the program (volunteers: 100 of 123; contractors: 351 of 450). Contrary to our hypothesis, the efficiency of ground-based shooting did not peak at sunset when sambar deer were most commonly detected with camera-traps. For both volunteer and contract shooters, CPUE was highest in the hours between sunset and midnight (Fig. 4), before declining and then peaking again before sunrise. CPUE was lowest during daylight hours.

Fig. 4.  Hourly catch per unit effort (CPUE; solid line) and effort (bars) for ground-based shooting of sambar deer by volunteer and contract shooters in Alpine National Park, Victoria, Australia, 2015–2020.

After accounting for all key expenses, the cost per hour for volunteer shooters was three to four times lower than that for contract shooters (Table 4). The contract shooters’ fees accounted for more than 75% of their cost per hour. As hypothesised, the higher effectiveness of contract shooters (i.e. less hours per deer killed than volunteer shooters) compensated for most of the cost per hour difference. However, across the 5-year program on both blocks, the cost per deer killed was higher for contract shooters than for volunteer shooters (12% in BHP and 10% in HWP; Table 4).

Table 4.  Cost-effectiveness of ground-based shooting of sambar deer for volunteer and contract shooters in Alpine National Park, Victoria, Australia, 2015–2020.

Empirical data collected during a 5-year management program showed that, given the same access opportunities and similar equipment, contract ground-based shooters killed four times more sambar deer per hour than did volunteer shooters in and around alpine peatlands in south-eastern Australia. This higher CPUE compensated for most, but not all, of the greater cost per unit effort of contract shooters. Hence, contract shooters were, on average, 10.1% more expensive per deer killed than were volunteer shooters. The CPUE and hence cost-effectiveness of volunteer shooters was strongly influenced by shooting method and was highest when shooting at night from a vehicle with a spotlight or when stalking with thermal-vision equipment. During the day, the use of gundogs to indicate deer significantly increased the CPUE of volunteer shooters. Both contract and volunteer shooters had a higher CPUE during the first half of the night and just before sunrise. Contractors and especially volunteer shooters relied on roads and tracks to move within the treatment units to shoot sambar deer close to alpine peatlands. Although contract shooters covered more area than did volunteer shooters, there were still large parts of the treatment units with little or no shooting effort after 5 years.

In our study area, the preference of sambar deer for denser wooded habitats and their crepuscular activity (Comte et al. 2022) is likely to explain why no sambar deer were killed by volunteer shooters during daylight stalking unless gundogs were used to detect deer. Gundogs have long been companions of deer hunters and have been associated with higher success in reducing overabundant mammalian species by ground-based shooters (89% and 47% success with and without dogs respectively; Bengsen et al. 2020). During our study, the highest CPUE was achieved at night, likely owing to deer feeding in more open areas where they are more easily detected by shooters using spotlights or thermal-vision equipment. Shooting from a vehicle with a spotlight has been associated with a decrease in sambar deer faecal pellet deposition in a flat open grassland adjacent to dense forest (Bennett et al. 2015). Even if using four-wheel-drive vehicles, this shooting method is restricted to the limited roads and tracks within the rugged study area. Using thermal-vision equipment should provide more shooting opportunities owing to improved detection rates compared to walking with binoculars during the day (Logan et al. 2019) or spotlights from vehicles during the night (Focardi et al. 2001).

On the basis of the results from the first year of shooting in the Bogong High Plains block, subsequent ground-shooting operations for volunteers were scheduled at night and with access to thermal-vision equipment, which was the primary method used by the contractors when they started shooting. Hunting deer on Victorian public land at night is prohibited [Wildlife (Game) Regulations 2012], and hence most volunteer shooters would have had less experience of night-shooting methods than would the contract shooters. During the management program, the CPUE of volunteer shooters increased over time in the treatment block HWP. This result suggests that volunteer ground-based shooters learnt to use new equipment, as has been described in the United States for volunteer shooters during a peri-urban white-tailed deer management (Hygnstrom et al. 2011). The difficulty of walking at night with a firearm in a rugged and often wooded landscape probably explains why volunteer shooters relied more on roads and flat areas to move than did contract shooters. The presence of roads has commonly been associated with increased deer hunting activity and CPUE (Lebel et al. 2012; Pauli et al. 2019; Rowland et al. 2021). By moving faster and covering more ground during their shifts, contract shooters may have been able to create more shooting opportunities than did volunteer shooters. According to the six contract shooters in the program, the key to successfully shooting sambar deer was to target deer emerging from wooded areas into open areas at night. Hence, during their shifts, contract shooters typically covered multiple feeding grounds with low wooded vegetation cover (K. Stone, pers. comm.).

Another factor potentially contributing to the higher CPUE of contractors than of volunteer shooters is that only the former had access to sound suppressors. This equipment is commonly used by contractors in ground-based shooting of deer in Australia (Hampton et al. 2022) and is likely to increase their CPUE by reducing the recoil and minimising the flight behaviours of non-targeted animals, thus creating more shooting opportunities when shooting individuals within a group (MacCarthy et al. 2011; Williams et al. 2018).

There is much interest in involving recreational hunters as volunteer shooters in deer control programs in Australia (Boyle and Henderson 2007; Bengsen and Sparkes 2016; New South Wales Natural Resources Commission 2017; Parliament of Victoria Environment, Natural Resources and Regional Development Committee 2017; Game Services Tasmania 2022). The Alpine National Park deer control trial used a mix of volunteer and contract shooters. If all the ground-based shooting was conducted by volunteer shooters (i.e. same number of shooting hours), then the cost of the program would have been 62% lower, but there would have been a 55% reduction in deer killed (315 fewer deer shot). In contrast, if contract shooters were used with the same effort, then the costs would have been 11% higher for a 52% increase in deer killed (300 more deer shot). Alternatively, contract shooters could have achieved the same outcome (i.e. total number of deer killed) with a 45% reduction in effort (20 fewer operations) and a 5% increase in costs. Volunteer shooters, to achieve the same outcome, would cost 43% less but require an additional 62 operations (40% increase) over the 5 years. However, the latter might not be possible because of limited human resources to organise the operations. There might also be a shortage of volunteer shooters because there is a limited amount of time that most volunteers can contribute during the seasons in which deer are present in and around the alpine peatlands. This was already apparent during the Alpine National Park deer control trial, with volunteer shooters showing a higher motivation to participate in ground-based shooting operations in BHP than HWP because recreational hunting was prohibited in the former (E. Thomas, pers. obs.). Our results support the general findings of Bengsen et al. (2020) that contract shooters are more likely to achieve population reduction, and within a shorter time, than are volunteer ground-based shooters.

Cost-effectiveness is a major aspect of successful management of overabundant deer populations, but community support is also important (Kilpatrick et al. 2007). There will often be greater community support for lethal control of overabundant deer when local people can contribute to the program (Martínez-Jauregui et al. 2020). Involving local hunters as volunteer shooters in a deer control program can therefore provide benefits to those individuals (i.e. the experience of hunting in a new area and meeting new like-minded people, taking home meat and/or a trophy and contributing to conservation programs) and increase community support for the program (Parliament of Victoria Environment, Natural Resources and Regional Development Committee 2017). Future studies of volunteer shooting programs in Australia could assess the motivations and experiences of the volunteers.

The differences in cost-effectiveness between volunteer and contract shooters during the Alpine National Park deer control trial were within the range of those observed during a peri-urban white-tailed deer ground-shooting program in Bloomington (Minnesota, USA) during 1991–1994 (Doerr et al. 2001). There, the cost per deer killed for volunteer shooters (US$117 per deer in 1994; A$304 in 2021) was 3% lower than for park rangers (US$121 per deer in 1994; A$314 in 2021) but 8% higher than for conservation officers (US$108 per deer in 1994; A$280 in 2021). Both categories of professional shooters had a much higher CPUE (4.7 h-per-deer for conservation officers and 2.1 h per deer for park rangers) than did volunteer shooters (30.9 h per deer). The costs per hour reported in the United States were much lower than in our study (accounting for inflation and currency rates) for both the organisational costs (A$39 compared with A$57 in our study) and for the professional shooters’ fees (A$39 compared with A$470 in our study). These large differences reflect the difference in salary between the two countries but could also result from different costing methods (e.g. inclusion or exclusion of overhead costs, insurances, vehicles and equipment). This highlights the importance of reporting detailed costing tools for deer management programs because costs vary greatly with local socioeconomic context.

The CPUE of ground-based shooters is expected to decline at an increasingly rapid rate as deer density declines (Batcheler and Logan 1963; Van Deelen and Etter 2003; Masters et al. 2018). An independent index of deer abundance (camera-trap detections per month) is available for one of the four pairs of treatment and non-treatment units (Comte et al. 2022; see Materials and methods), but that treatment unit received the greatest ground shooting effort. After accounting for seasonal and environmental factors, the sambar deer detection rate at BHP-B (treatment unit) decreased by 14% annually (2015–2019) but increased by 7% annually at the non-treatment unit BHP-C (Comte et al. 2022). In contrast to expectations, the apparent stability of the CPUE at both blocks during the program suggests that the ground shooting effort did not remove a sufficiently large proportion of the population to cause it to decline, but may have succeeded in locally preventing the general increase in sambar deer population observed at a larger scale across the eastern part of Victoria (Forsyth et al. 2018; Watter et al. 2020). Our analysis of the shooters’ GPS track-logs showed that the areas subject to intensive ground-shooting (combined UD50 for volunteers and contractors) represented only 10% (range 8–12%) of each treatment unit. This very localised shooting pressure may have meant that sambar deer killed by ground-based shooting may have been compensated for by local immigration (Porter et al. 2004; Putman 2012) from adjacent areas subject to no (BHP block) or less (HWP block) ground-based shooting. A management-scale experiment conducted in New Zealand to assess how forests responded to ground- and helicopter-based shooting of non-native deer was unable to consistently and substantially reduce deer densities at treatment relative to non-treatment areas (Forsyth et al. 2013), highlighting the difficulties of reducing deer populations in rugged and remote wooded landscapes.

If there is a limited time frame for deer control then contract shooters can be expected to remove more deer than volunteer shooters, but at a (slightly) greater cost. A limited time frame for control would occur when deer seasonally migrate outside the management boundaries (Jarnemo 2008), when frequent adverse climatic conditions limit the shooting opportunities (Baur et al. 2021), or when access to sites is temporally limited by health and safety concerns (e.g. areas used by the public; Reis and Higham 2009). If there is more than a small window of time to conduct the control, then involving local recreational hunters as volunteer shooters could decrease the cost of the program and increase community support for it.

The timing of the shooting and the supporting equipment that can be used greatly affects shooter effectiveness and hence cost. Ground-based shooting of deer is likely to be most effective when conducted at night (when deer are most active in open areas) and with thermal vision equipment (which maximise the probability of detecting deer without the shooter being detected by deer).

Last, ground-based shooting is likely to be most effective in areas with a high density of roads and tracks (Fraser and Sweetapple 1992; Rowland et al. 2021), which facilitate the movement of the shooters on foot and enable rapid movement around the landscape using vehicles.

Supplementary material is available online.

The data that support this study will be shared upon reasonable request to the corresponding author.

David Forsyth was a guest Associate Editor for this special issue. Despite this relationship, he did not at any stage have editor-level access to this manuscript while in peer review, as is the standard practice when handling manuscripts submitted by an editor to this journal. Wildlife Research encourages its editors to publish in the journal and they are kept totally separate from the decision-making process for their manuscripts. The authors have no other conflicts of interest to declare.

The Victorian Alpine Peatland Protection Program is an initiative jointly funded through Parks Victoria, the West Gippsland Catchment Management Authority, the Australian Government’s National Landcare Programme (via North East, West Gippsland and East Gippsland Catchment Management Authorities) and the Victorian Government through the Alps Intensive Management Program. The analysis of these data, and the writing of the manuscript, were funded by the Centre for Invasive Species Solutions project ‘Cost-effective management of wild deer’ (PO1-L-001).