SG: The stochastic shielding heuristic in ecological networks

要对生态资源做出正确的决策，就需要了解生态系统中某一部分的变化如何影响附近和远处的重要物种。发展一个关于生态系统变化的理论因几个因素而变得复杂。动物和植物物种与各种其他物种相互作用，并从生态系统的一个位置迁移到另一个位置。动植物物种的变化包括可预测的因素，如季节性迁徙，以及其他不可预测的因素，如每周和每天的温度和降雨量的变化。在这个项目中，研究人员开发了新的理论方法来理解不可预测的变化如何影响复杂的生态系统，为管理者提供了一个更好的框架来做出影响社会和学生的决策。生态系统中的物种以各种直接或间接的方式相互影响，从而通过生态网络连接起来。在当前的生态网络数学模型中，为了简化计算，忽略了不可预测的事件（称为随机性）。在这个项目中，研究人员将发展生态网络中随机过程的理论，为何时以及如何将随机波动纳入计算模型提供指导。具有现实复杂性（大量节点和连接）和人口规模（有限而不是无限）的生态网络模型自然会由于人口随机性（出生率和死亡率的变化）和环境随机性（极端天气、地震和山体滑坡等不可预测事件）而表现出随机波动。平均种群行为的波动可以发挥重要作用，例如在灭绝事件或入侵中，但在现实生态网络中建模所有随机元素在概念上和数字上都是困难的。在理论神经科学的最新创新（“随机屏蔽”启发式）的基础上，研究人员将开发一个框架，该框架可以大大减少准确表示一组感兴趣的节点（例如代表焦点物种或关键栖息地斑块）波动所需的独立随机过程的数量。该项目将创建一个框架，不仅利用平均人口规模数据，还利用差异数据，通过发展理论基础，补充现有的努力，将生态科学带入大数据时代。

Making good decisions about ecological resources requires understanding how changes in one part of an ecosystem can affect important species both nearby and far away. Developing a theory on ecosystem change is complicated by several factors. Animal and plant species interact with a variety of other species, and migrate from one location in an ecosystem to another. The changes in plant and animal species involve factors that are predictable, like seasonal migrations, and other factors that are unpredictable, like week-to-week and day-to-day changes in temperature and rainfall. In this project, the investigators develop new theoretical approaches for understanding how unpredictable changes affect complex ecosystems, providing managers with a better framework for making decisions that affect society and students with training. Species in ecosystems affect one another in a variety of ways, directly and indirectly, and thereby are connected by an ecological network. In current mathematical models of ecological networks, unpredictable events (referred to as stochasticity), are omitted for simplicity, in order to make the computations tractable. In this project, the investigators will develop a theory of stochastic processes in ecological networks to provide guidelines for when and how stochastic fluctuations should be included in computational models. Ecological network models with realistic complexity (large numbers of nodes and connections) and population sizes (finite rather than infinite) naturally exhibit random fluctuations due to demographic stochasticity (variation in birth and death rates) and environmental stochasticity (unpredictable events such as extreme weather, earthquakes and landslides). Fluctuations around average population behavior can play an important role, for instance in extinction events or invasions, but modeling all stochastic elements in a realistic ecological network is conceptually and numerically taxing. Building on recent innovations in theoretical neuroscience (the "stochastic shielding" heuristic) the investigators will develop a framework that can greatly reduce the number of independent stochastic processes needed to accurately represent the fluctuations in a select set of nodes of interest (e.g. representing focal species or critical habitat patches). The project will create a framework for exploiting data on not only average population sizes but also variances, by developing the theoretical underpinnings that complement existing efforts to bring ecological science into the age of big data.