Dissertation Research: Revealing the spatial distribution of risk in animal groups

非技术性动物群体以其与环境互动的能力而闻名，这种能力是个人无法做到的，例如通过迁徙、集体智力或逃避捕食者。例如，捕食者攻击高度协调的鱼群几乎没有成功，因为群体中的信息传输是如此有效。拍摄和跟踪大群动物的技术挑战限制了我们对它们如何在躲避捕食者方面如此有效的理解，但计算机视觉和高速拍摄的最新进展现在让我们能够准确地重建这些信息传输网络，并展示信息如何在动物群中流动。对这些网络的洞察提出了与长期确立的关于动物群体中哪里最安全的假设相左的预测：经典的行为生态学理论认为动物是中心，而新的网络数据表明群体的边缘。在这里进行的研究将首次检验这些相互竞争的假说。一只北方梭鱼捕食者和成群的金光鱼之间的现场互动将在实验室被拍摄下来，然后使用复杂的计算机视觉软件从感官网络的角度重建，以更好地了解信息是如何传输的，以及个体在群体中的位置的影响。该实验将提供对行为生态学、感觉生态学、博弈论和网络科学领域的见解。它将为研究生和本科生提供科学和统计培训，研究结果将在科学会议上传播，通过博客，并通过向当地高中进行科学推广。技术摘要几十年来，汉密尔顿的自私羊群理论一直作为对动物群体捕食风险空间分布的预期。在这个模型中，当动物移动到将其他个体置于自己和潜在的隐藏捕食者之间时，就会出现凝聚力很强的群体，从而最小化它们的“危险领域”。然而，对于鱼群等流动动物群体来说，对这一理论的支持褒贬不一，因为自私的群体理论不允许猎物对捕食者做出反应。另一方面，真正的捕食者和猎物之间的相互作用是动态的。直到最近，由于行为数据的质量和数量的技术限制，直接在鱼群中测试汉密尔顿的自私羊群理论是不可能的。然而，计算机视觉和高速摄像机的最新进展现在可以准确地测量鱼群所有成员的精细移动，以及对它们可获得的视觉信息的估计。根据这种视觉信息构建的网络可以准确预测运动决策是如何通过鱼群转移的。在这里，研究将通过在实验室拍摄金光鱼和北梭鱼之间的捕食者与猎物的相互作用，直接测试空间位置或网络结构是否是死亡风险更准确的预测因素的非排他性假设。

Non-technical AbstractAnimal groups are remarkable for their ability to interact with the environment in a way that individuals are unable to, such as through migration, collective intelligence, or predator avoidance. For example, predators attacking highly coordinated fish schools have little success due to how efficient information transmission is in the group. The technological challenges of filming and tracking large groups of animals have limited our understanding of how they are so effective at predator avoidance, but recent advances in computer vision and high-speed filming now allow us to accurately recreate these information transmission networks and show how information moves through animal groups. Insights into these networks have put forth predictions that are at odds with long-established hypotheses on where in the group animals are safest: classic behavioral ecology theory suggests the center, while new network data suggests the edge of the group. The research carried out here will for the first time test these competing hypotheses. Live interactions between a northern pike predator and schools of golden shiners will be filmed in the laboratory and then recreated from a sensory network perspective using sophisticated computer vision software to better understand how information is transferred and the ramifications of individual location in the group. The experiment will provide insights into the fields of behavioral ecology, sensory ecology, game theory, and network science. It will provide scientific and statistical training to graduate and undergraduate students, and findings will be disseminated at scientific conferences, through blogging, and through science outreach to local high schools.Technical AbstractFor decades, Hamilton's Selfish Herd theory has served as the expectation for the spatial distribution of predation risk in animal groups. In this model, cohesive grouping emerges as animals move to position other individuals between themselves and a potential hidden predator, hence minimizing their "domain of danger". Support for this theory has been mixed for mobile animal groups such as fish schools, however, because the Selfish Herd theory does not allow for prey to respond to the predator. Real predator-prey interactions, on the other hand, are dynamic. Until very recently, directly testing Hamilton's Selfish Herd theory in fish schools has been impossible due to technological limitations on the quality and quantity of behavior data. Recent advances in computer vision and high-speed cameras, however, now allow for accurate measures of the fine-scale movements of all members of fish schools, as well as estimations of the visual information available to them. Networks constructed from this visual information are an accurate predictor of how movement decisions transfer through schools of fish. Here research will directly test the non-exclusive hypotheses of whether spatial positioning or network structure is a more accurate predictor of mortality risk by filming predator-prey interactions between golden shiners and a northern pike predator in the laboratory.