arxiv: 2605.09064 · v1 · submitted 2026-05-09 · 📊 stat.AP

Recognition: 2 theorem links

· Lean Theorem

Bayesian decision theory for wildlife management under uncertainty: from inference to action

Abby Keller, Cyril Milleret, Olivier Gimenez

Pith reviewed 2026-05-12 02:10 UTC · model grok-4.3

classification 📊 stat.AP

keywords Bayesian decision theorywildlife managementuncertainty propagationutility functionswolf populationinvasive species controlecological modelingmanagement actions

0 comments

The pith

Bayesian decision theory extends standard inference into explicit wildlife management actions by evaluating utilities of alternatives under uncertainty.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows how to move from Bayesian models of population dynamics directly to choosing management actions such as wolf culls or muskrat control efforts. It does this by calculating the expected utility of each possible action from the full posterior distribution of model parameters, thereby folding uncertainty and trade-offs into the decision itself. A sympathetic reader would care because most ecological studies stop at parameter estimates or predictions, leaving managers to bridge the gap to real decisions without a formal way to weigh conflicting goals like species protection versus economic damage. The two worked examples demonstrate that the optimal choice is rarely the most extreme action but a compromise that accounts for how uncertainty affects each objective.

Core claim

Bayesian decision theory supplies a coherent workflow that treats management as the selection of an action maximizing expected utility, where utility is computed by averaging the consequences of each action over draws from the posterior distribution of the system. In the French wolf case this yields a recommended harvest level that balances the benefit of removals against the risk of driving the population too low. In the Dutch muskrat case the same procedure allocates a fixed control budget unevenly across provinces according to the relative weight placed on population reduction.

What carries the argument

Expected utility computed by averaging the value of each candidate action over posterior simulations of population trajectories, using explicit utility functions that encode the relative importance of ecological, economic and social outcomes.

If this is right

Optimal wolf harvest emerges as a compromise that reduces population risk while still achieving some removal benefit.
Optimal muskrat control effort rises with the priority given to population reduction and is distributed unevenly across space.
The workflow is a direct extension of existing Bayesian inference code rather than a separate modeling step.
Explicit utility functions make the trade-offs between objectives transparent and reproducible for managers and stakeholders.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same posterior-to-utility pipeline could be used for other repeated management decisions such as fisheries quotas or invasive plant removal schedules.
Once utilities are written down, sensitivity analyses can test how much the recommended action changes when different stakeholder groups supply their own weighting schemes.
The approach naturally produces a set of near-optimal actions rather than a single number, which may help negotiations when perfect consensus on objectives is impossible.

Load-bearing premise

That utility functions can be defined to accurately and acceptably capture complex, often conflicting ecological, economic, and social objectives without introducing arbitrary or post-hoc weightings that drive the optimal decision.

What would settle it

A demonstration that two different but plausible utility functions, when applied to the same posterior samples in either case study, produce substantially different recommended actions.

Figures

Figures reproduced from arXiv: 2605.09064 by Abby Keller, Cyril Milleret, Olivier Gimenez.

**Figure 1.** Figure 1: Bayesian decision theory workflow. A statistical model links data 𝑦 to ecological processes through a likelihood and a prior with parameter 𝜃. These are combined to derive the posterior distribution 𝑝(𝜃 ∣ 𝑦) ∝ 𝑝(𝑦 ∣ 𝜃)𝑝(𝜃). A set of candidate actions 𝑎 is defined. A utility function 𝑈(𝜃, 𝑎) specifies the consequences of each action. Expected utility E[𝑈(𝑎)] is computed by averaging over the posterior distr… view at source ↗

**Figure 2.** Figure 2: Bayesian decision framework for wolf management in France. (a) Temporal dynamics of wolf population size and removals. Blue points and vertical lines show capture–recapture estimates of population size with associated uncertainty, orange bars represent the number of wolves removed each year, and the dashed black line with shaded area corresponds to posterior mean and 95% credible intervals of the latent po… view at source ↗

**Figure 3.** Figure 3: Bayesian decision framework for muskrat management in the Netherlands. (a) Posterior predictions of total removals over time (year, month). Observed catches (blue points) are compared with model predictions (black dashed line) and their 95% credible intervals (shaded area). Trapping effort (orange bars) is shown on a secondary axis after rescaling. (b) Expected utility as a function of uniform trapping ef… view at source ↗

read the original abstract

Ecologists are increasingly expected to inform management decisions under uncertainty, yet most analytical workflows stop at statistical inference. This disconnect limits the practical impact of ecological modelling, particularly in high-stakes contexts such as wildlife management, where decisions must balance ecological, economic and social objectives. Bayesian decision theory provides a coherent framework to bridge this gap. It propagates uncertainty from posterior distributions to quantify the consequences of alternative actions through utility functions. Despite its strong theoretical foundations, it remains underused in ecology. Here, we present a practical workflow for implementing Bayesian decision theory using standard Bayesian tools. We illustrate the approach with two case studies. First, wolf management in France, where the decision consists of selecting the number of wolves that can be removed under uncertainty about population dynamics. Second, invasive muskrat management in the Netherlands, where the decision involves allocating a fixed control effort across space. In both cases, expected utility is computed from posterior simulations, explicitly accounting for uncertainty and trade-offs. Results show that optimal decisions emerge as a compromise between competing objectives. In the wolf case, optimal harvest balances removal benefits and population risk. In the muskrat case, optimal effort increases with the importance of population reduction and is unevenly allocated across provinces. These examples show that Bayesian decision theory can be implemented as a direct extension of standard inference. By making trade-offs explicit, it enhances transparency, reproducibility, and relevance for management. More broadly, it provides a flexible basis for integrating ecological modelling with decision-making.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper shows how to feed posterior draws into expected utilities for two wildlife decisions, but the utilities themselves are the part that needs more showing.

read the letter

This paper takes standard Bayesian posteriors from ecological models and uses them to pick management actions in two cases: how many wolves to remove in France under population uncertainty, and how to allocate fixed control effort for invasive muskrats across Dutch provinces. The workflow is presented as a direct next step after fitting the models, with optimal choices emerging from averaging utilities over posterior simulations rather than point estimates.

Referee Report

2 major / 2 minor

Summary. The manuscript claims that Bayesian decision theory provides a practical, direct extension of standard Bayesian inference for wildlife management under uncertainty. It propagates posterior uncertainty through utility functions to compute expected utilities and identify optimal actions, illustrated via two case studies: selecting the number of wolves to remove in France (balancing removal benefits against population risk) and allocating fixed control effort for invasive muskrats in the Netherlands (balancing effort against population reduction). The authors conclude that optimal decisions emerge as explicit compromises between objectives, enhancing transparency, reproducibility, and management relevance.

Significance. If the utility functions are fully specified with elicitation protocols and shown to be robust, the work could meaningfully advance the integration of ecological modeling with decision-making by making trade-offs explicit and accounting for uncertainty in high-stakes contexts. The use of standard Bayesian tools and real case studies is a strength for accessibility and demonstration of applicability.

major comments (2)

[Wolf management case study] Wolf management case study: the utility function encoding trade-offs between removal benefits and population risk is described only qualitatively, with no functional form, parameter values, elicitation protocol, or sensitivity analysis provided. This is load-bearing for the central claim that optimal decisions are a direct extension of the posteriors, as the reported compromise could be an artifact of arbitrary weightings rather than emerging from the posterior simulations alone.
[Muskrat management case study] Muskrat management case study: the utility trading off control effort against population reduction lacks any mathematical specification or details on how 'importance of population reduction' is quantified or propagated; without this, the claim that optimal effort 'increases with the importance' and is 'unevenly allocated across provinces' cannot be verified as reproducible from the posteriors.

minor comments (2)

The abstract and methods would benefit from a brief outline of the specific Bayesian models (e.g., population dynamics priors and likelihoods) used to generate the posteriors in each case.
Consider adding an appendix or table with example posterior draws, computed expected utilities for a range of actions, and the resulting optimal decisions to support reproducibility claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for their constructive feedback, which identifies key areas where greater explicitness is needed to strengthen the demonstration that Bayesian decision theory extends standard inference to actionable management. We address each major comment below and will revise the manuscript to incorporate the requested details, thereby improving reproducibility and supporting the central claims.

read point-by-point responses

Referee: [Wolf management case study] Wolf management case study: the utility function encoding trade-offs between removal benefits and population risk is described only qualitatively, with no functional form, parameter values, elicitation protocol, or sensitivity analysis provided. This is load-bearing for the central claim that optimal decisions are a direct extension of the posteriors, as the reported compromise could be an artifact of arbitrary weightings rather than emerging from the posterior simulations alone.

Authors: We agree that the wolf case study utility function requires explicit specification to substantiate the claim. In the revised manuscript we will provide the mathematical form of the utility function (a weighted combination of removal benefit and population risk terms), the specific parameter values used, the basis for those values (drawn from management literature and expert-derived thresholds for population viability), and a sensitivity analysis varying the weights to confirm that the optimal harvest level remains a direct output of the posterior simulations rather than an artifact of arbitrary choices. These additions will be placed in the main text with supporting code in the supplement. revision: yes
Referee: [Muskrat management case study] Muskrat management case study: the utility trading off control effort against population reduction lacks any mathematical specification or details on how 'importance of population reduction' is quantified or propagated; without this, the claim that optimal effort 'increases with the importance' and is 'unevenly allocated across provinces' cannot be verified as reproducible from the posteriors.

Authors: We accept this observation. The current description of the muskrat utility is insufficiently precise. In revision we will supply the explicit functional form (a linear or concave trade-off between total control effort and expected population reduction), define the 'importance of population reduction' as a scalar weighting parameter, and demonstrate its propagation through the posterior predictive simulations. We will also show analytically and numerically that optimal effort increases with this weight and is spatially uneven due to heterogeneity in the posterior distributions across provinces, with full reproducibility ensured via supplementary code. revision: yes

Circularity Check

0 steps flagged

Standard posterior-to-expected-utility workflow exhibits no circular reduction

full rationale

The paper presents Bayesian decision theory as a workflow that propagates posterior samples through utility functions to obtain expected utilities and optimal actions. No equations, fitted parameters, or self-citations are shown that would make the reported optimal decisions (wolf harvest levels or muskrat effort allocation) equivalent by construction to quantities already defined in the inference step. Utility functions are described as encoding explicit trade-offs but are not derived from or fitted to the same data in a self-referential manner. The central claim of a 'direct extension of standard inference' therefore rests on independent application of existing Bayesian machinery rather than on any load-bearing self-definition or renaming of results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard Bayesian inference producing usable posteriors and on the ability to specify utility functions that reflect real trade-offs; both are domain assumptions rather than new derivations.

free parameters (1)

utility function weights
Weights balancing ecological, economic, and social objectives must be chosen or elicited; these directly affect which action maximizes expected utility.

axioms (1)

domain assumption Posterior distributions from Bayesian models accurately represent all relevant uncertainty for decision making
Invoked when propagating posteriors to expected utilities in both case studies.

pith-pipeline@v0.9.0 · 5566 in / 1249 out tokens · 40040 ms · 2026-05-12T02:10:34.670347+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
expected utility E[U(a)] = ∫ U(θ,a) p(θ|y) dθ ... approximated as (1/N) Σ U(θ(n),a)
IndisputableMonolith/Foundation/BranchSelection.lean branch_selection unclear
U(a) = (b − c)a − α P(N_{T+1} < N_min)

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

[1]

Berger, J. O. 1985. Statistical decision theory and B ayesian analysis. Springer-Verlag, New York

work page 1985
[2]

Bos, D., E. E. van Loon, E. Klop, and R. Ydenberg. 2020. A Large-Scale Experiment to Evaluate Control of Invasive Muskrats . Wildlife Society Bulletin 44:314--322

work page 2020
[3]

Converse, S. J., C. T. Moore, M. J. Folk, and M. C. Runge. 2013. A matter of tradeoffs: Reintroduction as a multiple objective decision. Journal of Wildlife Management 77:1145--1156

work page 2013
[4]

Pradel, R

Cubaynes, S., R. Pradel, R. Choquet, C. Duchamp, J.-M. Gaillard, J.-D. Lebreton, E. Marboutin, C. Miquel, A.-M. Reboulet, C. Poillot, P. Taberlet, and O. Gimenez. 2010. Importance of accounting for detection heterogeneity when estimating abundance: The case of French wolves. Conservation Biology 24:621--626

work page 2010
[5]

Turek, C

de Valpine, P., D. Turek, C. J. Paciorek, C. Anderson-Bergman, D. T. Lang, and R. Bodik. 2017. Programming With Models : Writing Statistical Algorithms for General Model Structures With NIMBLE . Journal of Computational and Graphical Statistics 26:403--413

work page 2017
[6]

M., and F

Dorazio, R. M., and F. A. Johnson. 2003. Bayesian inference and decision theory - A framework for decision making in natural resource management. Ecological Applications 13:556--563

work page 2003
[7]

Hemming, V., A. E. Camaclang, M. S. Adams, M. Burgman, K. Carbeck, J. Carwardine, I. Chadès, L. Chalifour, S. J. Converse, L. N. K. Davidson, G. E. Garrard, R. Finn, J. R. Fleri, J. Huard, H. J. Mayfield, E. M. Madden, I. Naujokaitis-Lewis, H. P. Possingham, L. Rumpff, M. C. Runge, D. Stewart, V. J. D. Tulloch, T. Walshe, and T. G. Martin. 2022. An introd...

work page 2022
[8]

T., and M

Hobbs, N. T., and M. B. Hooten. 2025. B ayesian M odels: A S tatistical P rimer for E cologists, 2nd E dition. Princeton University Press

work page 2025
[9]

Keeney, R. L. 1996. Value-focused thinking: A path to creative decisionmaking. Harvard University Press

work page 1996
[10]

Kéry, M., and J. A. Royle. 2020. A pplied H ierarchical M odeling in E cology: A nalysis of D istribution, A bundance and S pecies R ichness in R and BUGS , V olume T wo: D ynamic and A dvanced M odels. Academic Press, London, UK

work page 2020
[11]

Loon, E. E. van, D. Bos, C. J. van Hellenberg Hubar, and R. C. Ydenberg. 2017. A historical perspective on the effects of trapping and controlling the muskrat ( O ndatra zibethicus ) in the N etherlands. Pest Management Science 73:305--312

work page 2017
[12]

Chapron, I

Marescot, L., G. Chapron, I. Chadès, P. L. Fackler, C. Duchamp, E. Marboutin, and O. Gimenez. 2013. Complex decisions made simple: A primer on stochastic dynamic programming. Methods in Ecology and Evolution 4:872--884

work page 2013
[13]

Duchamp, S

Milleret, C., C. Duchamp, S. Bauduin, C. Kaerle, A. Pirog, G. Queney, and O. Gimenez. 2026. Estimating wolf population size in France using non-invasive genetic sampling and spatial capture-recapture models. Biological Conservation 317:111772

work page 2026
[14]

Duchamp, C

Pirog, A., C. Duchamp, C. Kaerle, C. Dufaure de Citres, S. Rousselot, J. Lavarec, and G. Queney. 2025. Standardization of a High-Quality Methodological Framework for Long-Term Genetic Monitoring of the French Wolf Population . Ecology and Evolution 15:e71345

work page 2025
[15]

Polasky, S., S. R. Carpenter, C. Folke, and B. Keeler. 2011. Decision-making under great uncertainty: Environmental management in an era of global change. Trends in Ecology & Evolution 26:398--404

work page 2011
[16]

A., and R

Royle, J. A., and R. M. Dorazio. 2008. H ierarchical M odeling and I nference in E cology: T he A nalysis of D ata from P opulations, M etapopulations and C ommunities. Academic Press. San Diego, California

work page 2008
[17]

Runge, M. C., S. J. Converse, J. E. Lyons, and D. R. Smith (Eds.). 2020. Structured Decision Making : Case Studies in Natural Resource Management . Johns Hopkins University Press, Baltimore, Maryland

work page 2020
[18]

Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and management of animal populations. Academic Press, San Diego, California, USA

work page 2002
[19]

J., and M

Williams, P. J., and M. B. Hooten. 2016. Combining statistical inference and decisions in ecology. Ecological Applications 26:1930--1942. CSLReferences document

work page 2016