IIBR Informatics: A generalized modeling framework for integrating multi-species data sources to estimate biodiversity processes

生物多样性与生态系统的健康和完整性有关，物种对生态系统功能的贡献各不相同。评估整个物种群落的状态和动态是至关重要的，而不仅仅是那些拥有大量数据的物种。该项目开发了“综合群落模型”，这是一个统计建模框架，可以同时使用多物种数据源来估计生物多样性的现状、趋势和动态。目标是通过对鸟类、小型哺乳动物和蝴蝶等动物群落的模拟和实证案例研究，为估计物种和群落过程创建一个灵活的基础设施，该基础设施可以结合多种物种的多种数据类型。物种分布、丰度和人口统计率的估算是科学理解生物多样性动态和群落对外部威胁反应的基础，为生物保护提供重要信息。综合群落模型的发展将使研究人员能够在生物多样性加速丧失的时代获得跨时空尺度的物种和群落的详细推论。该项目还为研究生和博士后提供分层统计建模方面的培训，并创建一个K-12外展模块，向中学生传授生物多样性保护知识。综合群落建模框架使用分层方法合并单物种集成模型（结合目标物种的多个数据源）和分层群落模型（仅从单个数据源估计多物种发生或丰度模式）。尽管最近在单物种综合模型和分层群落模型方面取得了进展，但这两种方法都有缺点：前者仅限于单一物种，而后者未能利用合并多种数据源和数据类型所带来的好处（例如，同时估计丰度和人口比率，增加时空覆盖）。通过弥合单物种综合模型和分层群落模型之间的差距，综合群落模型利用了两者的能力，克服了传统上对生物多样性参数的狭隘推断（在空间、时间和获得的信息方面）。建模框架使用每个不同的可用数据源，通过层次化的观察模型与联合似然联系在一起，为底层生物过程模型的各个组成部分提供信息。群落的生物过程模型可以从简单的（如物种发生的估计）到复杂的（如物种生存、繁殖和丰度的估计），并取决于分类群的生物学和可用数据的数量/类型。该项目推动了种群和群落生态学领域的发展，因为它允许科学家利用多种数据源（尽管采样方案和时空数据结构和数量存在差异），从而提高了物种水平动态和生物多样性指标（如丰富度、组成）的准确性和精度。该项目的结果将在https://ezipkin.github.io.This上公布，该奖项反映了美国国家科学基金会的法定使命，并通过基金会的智力价值和更广泛的影响审查标准进行评估，认为值得支持。

Biodiversity is linked to the health and integrity of ecosystems with species varying in their contributions to ecosystem functions. It is critical to assess the status and dynamics of whole communities of species and not just those species that have large amounts of data. This project develops ‘integrated community models’, a statistical modeling framework to simultaneously use multi-species data sources to estimate the status, trends, and dynamics of biodiversity. The objective is to create a flexible infrastructure for estimating species and community processes that can incorporate multiple data types on multiple species through simulations and empirical case studies on animal communities including birds, small mammals, and butterflies. Estimates of species distributions, abundances, and demographic rates form the basis of scientific understanding of biodiversity dynamics and community responses to external threats, delivering critical information for biological conservation. The development of integrated community models will enable researchers to obtain detailed inferences on species and communities across spatiotemporal scales during an era of accelerated biodiversity loss. This project also provides training to graduate students and postdoctoral scholars in hierarchical statistical modeling and creates a K-12 outreach module to teach middle school students about biolodiversity conservation.The integrated community modeling framework uses a hierarchical approach merging single-species integrated models (which combine multiple data sources on a target species) and hierarchical community models (which estimate multi-species occurrence or abundance patterns but only from a single data source). Although there have been recent advances in single-species integrated models and hierarchical community models, both approaches have shortcomings: the former is limited to a single species, whereas the latter fails to take advantage of the benefits gained from merging multiple data sources and data types (e.g. estimation of both abundance and demographic rates simultaneously, increased spatiotemporal coverage). By bridging the gap between single-species integrated models and hierarchical community models, integrated community models leverage the capabilities of both and overcome traditionally narrow inferences (in terms of space, time, and information gained) on biodiversity parameters. The modeling framework uses each of the different available data sources to inform various components of the underlying biological process model through hierarchical, observation models linked together with a joint likelihood. The biological process models for communities can range from simple (e.g. estimates of species occurrence) to complex (e.g. estimates of species survival, reproduction, and abundance) and depend on both the biology of the taxonomic group and the quantity/type of available data. This project advances the fields of population and community ecology because it allows scientists to take advantage of multiple data sources (despite differences in sampling protocols and spatiotemporal data structures and quantities), leading to increased accuracy and precision of species-level dynamics and biodiversity metrics (e.g. richness, composition). The results of the project will be made available at https://ezipkin.github.io.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.