Using stochastic dynamic programming for making decisions about invasive species

Abstract

Invasive species pose significant ecological, economic, and public health challenges. Their management requires repeated decisions under uncertainty about when, where and how intensively to intervene. Although structured decision-making and adaptive management provide conceptual frameworks for addressing these challenges, their practical implementation remains limited because optimization methods such as stochastic dynamic programming (SDP) are often perceived as mathematically complex, challenting to apply and difficult to interpret. Here, we provide a practical guide to using SDP and its extensions for invasive species management. Using coypu (Myocastor coypus) management as a running case study, we introduce the framework through a minimal working example before progressively extending it to spatial dynamics, imperfect detection using partially observable Markov decision processes (POMDPs), adaptive management under model uncertainty, and value-of-information analyses that quantify the benefits of learning before acting. All examples are accompanied by reproducible implementations in R, together with recommendations on model formulation, parameterization and interpretation. We also discuss the strengths and limitations of SDP. By lowering the technical barrier to SDP, this guide aims to help ecologists and wildlife managers integrate decision theory into invasive species management, enabling transparent, reproducible and objective-driven management strategies under uncertainty. More broadly, the methods presented here are applicable to a wide range of ecological decision problems beyond invasive species, wherever management requires balancing immediate actions against uncertain future outcomes.