SEES Fellows: Developing Semi-parametric Models, Algorithms, and Tools for Ecological Analysis of Species Biodiversity

该项目的目标是开发统计方法，使生态学家能够解决生物多样性建模中的关键问题。 这项研究将推进一个一般的半参数方法，适用于物种分布模型（SDM）的物种占有率，丰度，人口统计和动态。 该方法将解决空间数据模型面临的两个主要挑战：1）对物种实地观测中的漏测进行统计校正，以获得感兴趣的生态量的无偏估计，以及2）捕捉环境输入和模型变量之间的复杂关系。虽然SDMS中的当前技术包括解决第一个挑战的参数模型和解决第二个挑战的非参数方法，但所提出的方法将是第一个同时解决这两个问题的方法。 层次模型是一类重要的物种分布模型，它通常包含一些不可观测的生态变量。 这些模型具有生物学意义的参数（例如灭绝概率等）。这可以与描述各种环境特征（例如生境、土地使用、气候等）的一组输入变量相联系。 一旦拟合，可以检查模型以了解输入对参数的影响。 然而，这个拟合过程往往是困难的，因为需要仔细选择哪些输入包括在模型中，以及假设什么样的功能形式的影响。 将开发的方法将灵活的非参数方法纳入这些层次模型。 由此产生的方法将保留语义的层次模型，但允许输入的影响，以适应灵活，自动捕捉非线性和相互作用，而无需广泛的modelselection.Rapid下降，本土物种的栖息地是一个全球性的问题，由于越来越多的人类占主导地位的土地利用，栖息地碎片化，和气候变化。 为了设计保护区、保护地役权和类似的政策措施，需要了解受威胁物种在时间和空间上的栖息地要求和种群动态。 因此，需要精确的人口统计和物种分布模型。 该项目开发的方法将促进对物种分布关键方面的理解，从而指导保护工作。 建模方法将应用于各种生态框架，生态学合作者的反馈将有助于确保决策者的可用性。 此外，该项目的跨学科研究工作将在面向小学、初中和高中学生的外展计划中得到强调。（SEES研究员）计划，其目标是帮助实现所需的发现，为导致环境，能源和社会可持续性，同时创造必要的劳动力来应对这些挑战。 在SEES研究员的支持下，该项目将使一个有前途的早期职业研究人员能够在与可持续发展相关的独立研究生涯中建立自己的地位。

The goal of this project is to develop statistical methods to enable ecologists to address critical problems in biodiversity modeling. The research will advance a general semi-parametric methodology that applies to species distribution models (SDMs) of species occupancy, abundance, demographics, and dynamics. The approach will address two major challenges for SDMs: 1) statistically correcting for missed detections in field observations of the species to attain unbiased estimates of ecological quantities of interest, and 2) capturing complex relationships between environmental inputs and model variables. While current techniques in SDMs include parametric models that address the first challenge and nonparametric methods that address the second challenge, the proposed methodology will be the first to simultaneously address both. Hierarchical models are one important class of species distribution models, which often contain unobserved variables of ecological interest. These models have parameters with biological meaning (e.g. extinction probability, etc.) that can be linked to a set of input variables describing various environmental characteristics (e.g. habitat, land use, climate, etc.). Once fit, the models can be examined to understand the effects of the inputs on the parameters. However, this fitting process is often difficult due to the need to carefully select which inputs to include in the model and what to assume about the functional form of their effects. The methods to be developed will incorporate flexible nonparametric methods into these hierarchical models. The resulting methodology will retain the semantics of the hierarchical model but allow the input effects to be fitted flexibly, automatically capturing nonlinearities and interactions without extensive model selection.Rapid declines in habitat for native species are a global problem due to increasingly human-dominated land-use, habitat fragmentation, and climate change. To design reserves, conservation easements, and similar policy measures, there is a need to understand habitat requirements and population dynamics of threatened species across time and space. Hence, accurate models of the demographics and distribution of species are needed. The methodology to be developed by this project will advance understanding of key aspects of species distributions that can guide conservation efforts. Modeling methodologies will be applied to a variety of ecological frameworks, and feedback from collaborators in ecology will help ensure usability by decision-makers. Moreover, the interdisciplinary research efforts of this project will be highlighted in outreach programs to elementary, middle, and high school students.This project is supported under the NSF Science, Engineering and Education for Sustainability Fellows (SEES Fellows) program, with the goal of helping to enable discoveries needed to inform actions that lead to environmental, energy and societal sustainability while creating the necessary workforce to address these challenges. With SEES Fellows support, this project will enable a promising early career researcher to establish themselves in an independent research career related to sustainability.