🧬 biology

A Spatial Model of Chronic Wasting Disease and the Optimal Placement of a Sight-Based Cull

This study employs a spatial reaction–diffusion model on real landscapes to demonstrate that while an optimal sight-based cull of clinically infected cervids can slow the spread of Chronic Wasting Disease and recover susceptible populations, it is mathematically incapable of halting the invasion due to significant transmission from pre-clinical animals and environmental reservoirs.

Original authors: Gabe Flanagin, Alex Escoto, Zhen Chao

Published 2026-11-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Gabe Flanagin, Alex Escoto, Zhen Chao

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Chronic wasting disease is a fatal illness that affects deer, elk, and moose across North America and parts of Europe and Asia. Unlike many other wildlife diseases that spread only when animals touch each other, this illness has a second, invisible route of transmission. Infected animals shed microscopic infectious particles, called prions, in their saliva, urine, and feces long before they ever look sick. These particles bind to the soil and can remain infectious for years, creating a persistent environmental reservoir. Even if all the infected animals are removed from a field, the ground itself can still infect new deer that walk across it. Because there is no vaccine and no cure, wildlife managers rely on removing animals to slow the spread, but they have little guidance on where or when to focus these efforts. The disease moves slowly across the landscape, often taking decades to reach high levels, making it difficult to know if a management strategy is working until it is too late.

To solve this puzzle, researchers at Western Washington University built a detailed computer model that simulates how the disease moves across real terrain. Instead of treating the land as a flat, empty sheet, they used digital elevation data to create a three-dimensional surface that includes mountains, valleys, and different types of vegetation. In their model, the land itself dictates how many deer a place can support and how easily they can move through it. Steep slopes slow deer down, while certain types of land cover allow them to move more freely. The model tracks five distinct groups: healthy deer, those who are infected but not yet showing signs, those who are infected and shedding the disease but still look healthy, those who are visibly sick, and the invisible layer of prions waiting in the soil. The researchers then asked a specific question: if a field crew can only identify and remove the deer that are visibly sick, where and when should they do it to do the most good?

The team used advanced mathematical techniques to find the best possible strategy for this limited type of culling. They did not just guess; they calculated the exact outcome of every possible placement of effort over a twenty-year period. Their simulations revealed a counterintuitive truth about how to fight the disease. The most effective strategy is not to cull the deer where the sickness is currently most visible or where the disease was first found. Instead, the optimal effort forms a ring that moves ahead of the disease, targeting the leading edge of the invasion. In their simulations, this ring of removal effort positioned itself between 12 and 19 kilometers ahead of the peak of visibly sick animals. By removing the sick animals before they can reach a large population of healthy deer, the strategy protects the untouched herds waiting in the path of the disease.

Despite this precise targeting, the study found that a cull limited to visibly sick animals cannot stop the disease entirely. The researchers proved mathematically that because the disease spreads so heavily through the environment and through animals that look healthy, removing only the sick ones leaves a large "blind spot." Even with the maximum possible effort, the disease front continues to move forward, retaining between 56 and 94 percent of its original speed. The best that can be achieved is a slowing of the spread and a recovery of between 7 and 53 percent of the healthy deer population that would otherwise have been lost. The study also showed that simply counting the total number of new infections can be misleading. Because the cull keeps more deer alive, it can sometimes lead to a higher total number of infections over time, even though the risk to any single deer is actually lower.

The landscape itself plays a crucial role in how much effort is needed. When the researchers applied their model to three real landscapes in Colorado, they found that the terrain and land cover did not change the overall benefit of the strategy, but they did change where the effort should go. The disease naturally avoids steep, rugged ground and areas with poor vegetation. By letting the model account for these natural barriers, the researchers found that the same level of protection could be achieved with 26 to 35 percent less effort. The optimal strategy simply skips the difficult terrain where the disease is unlikely to spread, focusing resources only on the paths the deer actually take.

Ultimately, the study provides a clear, data-driven guide for wildlife managers, showing that while a sight-based cull cannot eliminate chronic wasting disease, it can significantly slow its advance if placed correctly. The most effective action is to target the front lines of the invasion rather than the center of the outbreak. However, the research also highlights the limits of this approach. Because the disease spreads through the soil and through animals that look healthy, managers cannot rely on culling alone to save a herd. The findings suggest that to truly close the gap, future efforts must also address the environmental reservoir and find ways to manage the invisible, pre-clinical stage of the infection.

Technical Summary: A Spatial Model of Chronic Wasting Disease and the Optimal Placement of a Sight-Based Cull

Problem Statement
Chronic wasting disease (CWD) is a fatal prion disease affecting cervids, characterized by transmission via direct contact and a persistent environmental reservoir. Current management relies heavily on culling, yet empirical evidence regarding the optimal spatial and temporal placement of these efforts is limited. Existing models often lack spatial explicitness, ignore landscape heterogeneity (terrain and land cover), or fail to address the specific constraint that field crews can only identify and cull clinically affected animals, missing the pre-clinical infectious stage. This paper addresses the question: Where and when should a sight-based cull be applied on a real landscape to maximize its efficacy, and what are the theoretical limits of such a strategy?

Methodology
The authors formulate a five-compartment reaction-diffusion model (Susceptible, Exposed, Pre-clinical, Clinical, and Environmental Prion Load) defined on a surface $\Gamma \subset \mathbb{R}^3$ derived from digital elevation data.

Spatial Dynamics: The model incorporates anisotropic diffusion tensors that redirect deer movement along terrain contours based on slope and land cover permeability. Carrying capacity is modulated by land cover quality.
Control Formulation: The study frames an optimal control problem where the control variable $v(x,t)$ represents a time- and space-dependent removal rate applied exclusively to the clinical class ( $R$ ). The objective functional minimizes the accumulated susceptible deficit, environmental prion load, and an elastic net penalty on effort (combining $L^2$ and $L^1$ norms to promote sparsity).
Numerical Implementation: The authors employ surface finite elements on a triangulated mesh. They derive the exact discrete adjoint of the semi-discrete IMEX (Implicit-Explicit) time-stepping scheme to compute gradients for optimization. This ensures the gradient is exact with respect to the discrete objective, avoiding the pitfalls of inexact gradients in non-convex problems. The optimization utilizes a limited-memory, bound-constrained quasi-Newton method (L-BFGS-B).
Parameterization: Parameters are drawn from literature (primarily mule deer data), with specific attention to transmission routes, shedding rates, and prion persistence. The study sweeps across environmental decay rates ( $\delta$ ) and basic reproduction numbers ( $R_0$ ) to test robustness.

Key Contributions

Spatially Explicit Model: The paper presents a continuous reaction-diffusion model on a realistic 3D terrain surface, integrating land cover into both carrying capacity and anisotropic diffusion, a feature often missing in previous CWD models.
Closed-Form Bound: The authors derive a model-independent, closed-form bound for the control reproduction number ( $R_c$ ) achievable by a sight-based cull. This bound demonstrates that such a cull cannot reduce transmission below replacement ( $R_c < 1$ ) if the pre-clinical transmission fraction is too high relative to the maximum feasible culling effort.
Exact Discrete Adjoint: The study implements and verifies (via Taylor tests) the exact discrete adjoint of a surface finite element scheme for a complex, multi-compartment epidemic system, enabling efficient gradient-based optimization on realistic landscapes.
Optimal Placement Strategy: Through factorial sweeps and ablation studies on three Colorado landscapes, the paper characterizes the spatial structure of the optimal cull.

Results

Limits of Control: Under plausible effort caps, a sight-based cull cannot halt the invasion or eliminate the disease. The pre-clinical stage generates approximately 39.5% of local transmission, creating a "blind spot" that the cull cannot address. Consequently, the control reproduction number remains above 1.36 even with maximal effort.
Invasion Dynamics: The optimal cull slows the invasion front but does not stop it. The front retains 56–94% of its uncontrolled speed. The control recovers 7–53% of the accumulated loss of susceptible deer, with efficacy decreasing as transmissibility ( $R_0$ ) and reservoir persistence increase.
Spatial Placement:
- In non-depleted herds, effort forms a disk around the infection focus.
- In depleted herds (high $R_0$ ), the optimal effort forms a ring at the leading edge of the invasion, positioned 12–19 km ahead of the clinical peak. This placement targets the intact susceptible herd that the next clinical animal would infect, rather than the current clinical peak.
- Cumulative incidence may rise under control because the cull keeps more susceptible deer alive, exposing them to the reservoir; thus, per-capita risk metrics are more appropriate than cumulative incidence.
Landscape Impact: While terrain and land cover significantly alter where effort is applied (e.g., following valley corridors), they have a minimal effect on the total benefit (deficit reduction). However, the full model (with terrain and land cover) achieves the same benefit with 26–35% less total effort compared to isotropic, homogeneous models, primarily by avoiding culling in areas where the invasion does not penetrate.

Significance and Claims
The paper claims that while a sight-based cull is insufficient to eradicate CWD due to the significant transmission contribution of the pre-clinical stage, it can effectively slow the spatial spread and reduce the total burden on the herd. The study emphasizes that optimal management requires targeting the leading edge of the invasion rather than the clinical epicenter. Furthermore, it highlights that landscape heterogeneity allows for more efficient resource allocation, reducing the total effort required without sacrificing the reduction in susceptible deficit. The authors conclude that the deposition ratio (pre-clinical vs. clinical shedding) and the route ratio (environmental vs. direct transmission) are the critical parameters for field estimation to refine these conclusions, and they suggest that carcass removal and environmental decontamination are necessary levers to close the gap left by sight-based culling.

More like this