Managing large-scale invasions: Simulation-optimization with Gaussian dispersal Kernels and stochastic seed establishment
本文首次将入侵者随机出现（非生物物理影响）整合到模拟优化模型中，用于控制Sericea lespedeza，通过预测性高斯种子扩散与随机发芽算法捕捉传统模型遗漏的长距离建立事件，并在预算约束下确定最优搜索和处理位置。
Biological invaders, such as Sericea lespedeza, cause over $21 billion (about $65 per person in the US) in annual losses for the US, necessitating effective control methods. To our knowledge, this article is the first to integrate random occurrences of an invader that are not attributable to biophysical impacts, within an integrated simulation-optimisation model to control Sericea. Specifically, we introduce a novel dispersal framework that integrates predictable Gaussian seed spread with a random sprout algorithm, explicitly addressing the long-standing question of random pop-ups of new invaders and capturing long-distance establishment events that traditional models miss. The simulation models the species’ biological growth and integrates both predictable dispersal and unpredictable establishment events into a unified framework. Our optimisation model minimises economic damage by determining optimal search and treatment locations under budget constraints. The case study data and parameter calibration are based on large-scale field data collected in Kansas and Oklahoma. We simulate Sericea growth over a 2,500-acre landscape for 25 years, representing a 25-fold increase in spatial coverage and more than double the temporal scope compared to former studies, substantially increasing problem complexity while demonstrating the scalability of our model. Results, averaged over 10 independent replications, show that prioritising searches in low-density areas and treating infestations immediately upon detection yield the greatest benefits. The framework highlights the value of early detection, search speed, and cost-effective control, offering a generalisable tool for invasive species management.