Hunting strategies to increase detection of chronic wasting disease in cervids

Culling is a common but contentious tool for controlling wildlife disease outbreaks and can conflict with conservation objectives. Early action is crucial for diseases that develop environmental reservoirs, such as anthrax and chronic wasting disease (CWD). For infections with latent stages and low early prevalence, livestock management may adopt preemptive culling, but this is often politically infeasible for wildlife of conservation concern. The study develops a general approach termed proactive hunting surveillance to detect wildlife disease in early epidemic stages without causing substantial population declines. The approach targets demographic groups with higher infection probability and lower reproductive value, integrating selective harvesting with veterinary epidemiology concepts of freedom from infection and risk-based surveillance. The method is applied to CWD in Norwegian wild reindeer, where infection was eradicated in one population but the status of adjacent populations remained uncertain, and where preemptive culling is undesirable due to conservation concerns.

The paper situates its contribution within research on culling as disease control, emphasizing the controversy in wildlife (e.g., badger culling for TB) and the increasing attention to wildlife diseases such as CWD and African swine fever. Prior CWD surveillance often relies on testing hunter-harvested cervids and requires large sample sizes for early detection, with limited strategic selectivity. Demographic patterns of infection are common in wildlife, and CWD typically shows higher prevalence in adults than in juveniles and often higher in adult males than females in several cervids. Population dynamics literature indicates males in polygynous species are usually not limiting for population growth unless sex ratios are extreme, supporting targeted male harvest to minimize demographic impacts. The study draws on risk-based surveillance and freedom-from-disease frameworks in veterinary epidemiology, advocating use of a design prevalence for documenting freedom. It also references spatial sampling importance in North American cervids and contrasts with more discrete alpine reindeer populations, which facilitate two-level sampling strategies akin to livestock settings.

Study areas: Two Norwegian alpine reindeer populations were analyzed—Nordfjella zone 2 (~1000 km²), adjacent to the previously infected and eradicated Nordfjella zone 1, and Hardangervidda (~8000 km²), the largest wild reindeer population. Hunting seasons were extended to meet surveillance goals, and marksmen conducted additional culling in Nordfjella zone 2 in early 2019. Modeling framework: A three-component pipeline linked (1) a population estimation model, (2) a population simulation model, and (3) a disease detection model, with outputs used iteratively to estimate surveillance sensitivity and probability of freedom from infection over time. 1) Population estimation model: Using four annual surveys (winter aerial minimum counts; mid-summer calving surveys; harvest records; post-hunt rutting ground counts of sex/age composition), the authors fitted a hierarchical change-in-ratio model (Bayesian inference via rjags) to estimate initial abundance, sex/age structure (calves, yearlings, adults), and demographic rates (stochastic recruitment and calf summer survival; constant winter survival). Uninformative priors were used; three MCMC chains of 250,000 iterations with burn-in and thinning assessed convergence. Outputs for 2018 fed the simulation model. 2) Population simulation model: A stochastic, stage-structured, two-sex matrix model (calves, yearlings, adults) projected populations under harvest. Key parameters included carrying capacity (K), minimum threshold for adult females, and target post-harvest adult sex ratio (male:female) thresholds (1:3, 1:5, 1:10, 1:20). Harvest strategies: - Ordinary: sex- and age-specific harvest rates set to historical 3-year means, with constraints to avoid dropping below sex ratio and female thresholds. - Proactive: adjust adult male harvest to achieve specified post-harvest sex ratio while stabilizing population below K by tuning adult female harvest rate; constraints enforced minimum adult females. Stochasticity: demographic rates and initial sizes were drawn from specified distributions; 1000 iterations per scenario. Outputs per year included numbers harvested and post-harvest population size/structure. 3) Disease detection model: A stochastic scenario-tree model estimated diagnostic sensitivity per sampled individual as a function of age, sex, time since infection, sample type and quality, and ELISA test sensitivity dynamics for retropharyngeal lymph nodes and brain tissue. Infected individuals were assigned random times since infection; calves were excluded due to very low detectability. Relative risk (RR) of infection by demographic group used a baseline of 1:2:6 (yearlings:adult females:adult males), with alternatives (1:1:1, 1:2:2, 1:2:4) explored. The model computed annual surveillance sensitivity (SSe) at a specified design prevalence (set as a number of infected individuals among yearlings + adults) using a hypergeometric approximation, assuming all harvested animals were tested. The probability of freedom from infection each year was updated via Bayes theorem using SSe and an annual probability of introduction (scenarios: 5% for Nordfjella zone 2 and 1% for Hardangervidda before eradication of zone 1; 0.1% after eradication; also a constant high-risk scenario explored). Initial prior infection probability was 0.5. For each harvest strategy and epidemiological parameter combination, 1000 iterations produced distributions of SSe and freedom-from-infection probabilities. Empirical implementation: The model incorporated observed 2016–2019 harvest/testing and simulated 2019–2029 trajectories. Ordinary harvest rates by class were taken from 2016–2018 means. The approach quantified time to reach target probabilities (e.g., 90% and 99%) of freedom from infection at specified design prevalence thresholds.

• Proactive hunting surveillance (targeted, male-biased harvest while maintaining female numbers) substantially reduced time to document freedom from CWD compared to ordinary harvest surveillance. • Headline result: Achieved 99% probability of freedom from infection at a design prevalence threshold of fewer than 4 infected reindeer within 3–5 years under proactive surveillance, versus roughly 10 years under ordinary harvest surveillance. Nordfjella zone 2 (smaller population): • Ordinary harvest (calves 7.4%, yearlings 7.7%, adult females 11.3%, adult males 14.3%): ~6 years (2018–2023) to 90% and ~11 years (2018–2028) to 99% probability of freedom at RR 1:2:6. • Optimal proactive strategy: cull all available adult males in year 1 and harvest just enough females to maintain population within thresholds; this stabilized female numbers while accelerating detection. • Sex ratio thresholds (male:female) after harvest markedly affected timelines. For 1:3, 1:10, 1:20 thresholds, time to 90% was 3, 2, and 2 years; time to 99% at least 5, 4, and 3 years, respectively. • Continuing calf/yearling harvests had minimal effect on freedom-from-infection probabilities after 5 years; male-biased adult harvest added information via higher RR and greater sample sizes without reducing female population. Hardangervidda (larger population): • Ordinary harvest (calves 13.9%, yearlings 15.8%, adult females 14.3%, adult males 18.0%): ~4 years (2018–2021) to 90% and ~10 years (2018–2027) to 99% probability of freedom. • With sex ratio thresholds 1:3, 1:10, 1:20: time to 90% was 2, 1, and 1 years; time to 99% at least 5, 4, and 3 years, respectively, with no decline in female numbers. Epidemiological sensitivities: • Relative risk pattern strongly influenced outcomes when harvest was male-biased; under proactive (adults only, cull adult males to m:f=1:5 in 2018), probability of freedom increased with stronger male risk bias (e.g., Nordfjella zone 2: ~70% for RR 1:1:1; 73% for 1:2:2; 84% for 1:2:6). • Design prevalence critically affected time to reach target confidence; lower design prevalence (stricter standard) required longer surveillance. • Higher annual probability of introduction (e.g., if adjacent infected population remained) delayed attainment of high confidence. Empirical surveillance outcomes: • Nordfjella zone 2 2018: ~63% (95% CI 62–65%) freedom from infection. Additional marksmen cull of 50 adult males + 2 adult females in winter 2019 raised this to ~75% (73–77%). After ordinary 2019 hunting, increased to ~86% (84–88%). • Hardangervidda 2018: ~68% (67–69%) freedom from infection; planned 2019 increase to ~1170 adult males harvested raised certainty to ~86% (84–87%).

Targeted, male-biased harvest leverages typical CWD demographic patterns (higher prevalence in adults, often higher in males) and polygynous mating systems wherein males are not usually limiting for population growth. By concentrating sampling on higher-risk, lower-reproductive-value classes, proactive hunting surveillance accelerates early detection (or substantiation of freedom) while minimizing demographic impacts, notably preserving adult female numbers. The approach reduced time to 90% and 99% confidence by several years compared to ordinary harvest surveillance in both study populations. The design prevalence selection is a managerial decision with major implications: a lower threshold (e.g., fixed small number of infected individuals) ensures earlier detection in absolute terms but necessitates more intensive or prolonged surveillance to reach high confidence of freedom. Spatial population structure and proximity to infection sources inform the annual introduction risk, which can substantially influence surveillance outcomes; discrete alpine reindeer populations allow adaptation of two-level sampling strategies akin to livestock management. While effective, targeted male harvest may alter rut timing and synchrony and raises ethical and social concerns (e.g., orphaning if females are taken without calf harvest). Implementation challenges include hunter acceptance, quota fulfillment, and conflict; marksmen can improve effectiveness but may heighten social controversy. Participatory modeling with stakeholders can facilitate adoption by explaining the epidemiological logic and balancing conservation with disease prevention.

The study introduces and demonstrates proactive hunting surveillance as a practical alternative to preemptive culling for early detection and freedom-from-infection substantiation in wildlife. By strategically biasing harvest toward demographic groups with higher infection risk and lower reproductive value, managers can markedly shorten the time to high confidence of CWD absence while maintaining female population stability. Applied to Norwegian reindeer, the approach achieved 99% confidence within 3–5 years under realistic constraints versus about a decade with ordinary surveillance. The framework integrates selective harvesting with risk-based surveillance and Bayesian updating and is adaptable to other wildlife diseases given clear demographic infection patterns, selective harvest feasibility, and basic population modeling capacity. Future work should refine epidemiological parameterization (e.g., introduction probabilities, dynamic relative risks), evaluate long-term demographic side effects, and enhance stakeholder engagement strategies for practical implementation.

Key parameters, including relative risk patterns and probabilities of introduction, relied on expert judgment due to limited empirical data. The model assumes a constant demographic infection pattern appropriate for early epidemic stages but may not capture temporal shifts. Calves were excluded from detection due to low test sensitivity, potentially omitting rare early infections. Test sensitivity functions and scenario-tree pathways depend on available validation data for tissues and ELISA performance. The design prevalence choice strongly affects timelines and reflects management preferences rather than biological thresholds. Social and implementation uncertainties (hunter behavior, quota fulfillment, conflicts over marksmen use, ethical concerns about harvesting females with offspring) can affect realized sampling and outcomes. Population dynamic effects of skewed sex ratios (e.g., altered rut timing/synchrony) and potential movements by females seeking mates could introduce unmodeled risks.