Postdoctoral Fellowship: OPP-PRF: Leveraging Community Structure Data and Machine Learning Techniques to Improve Microbial Functional Diversity in an Arctic Ocean Ecosystem Model

北冰洋正在经历迅速的环境变化，对极地海洋生态系统的几乎所有方面都产生了连锁影响，包括海洋微生物、微生物(即细菌和古生物)的丰度和群落组成，这些微生物在冰-海系统内的元素循环中发挥关键作用。数值模拟是测试生态机制和预测变化影响的关键方法。然而，目前的方法通常没有很好地解决细菌多样性或活动率问题，这是未来预测关键海洋生态系统功能(如生物碳减少)的不确定性来源。通过对当前北极微生物群落序列观测的统计探索，本研究旨在将我们对细胞环境响应的理解转变为与生态系统过程相关的尺度，并改进微生物群落结构和功能多样性的数值模拟。该项目资助一名博士后学者，并支持本科生培训，以建设北极计算研究的能力。利用来自北冰洋中部的公开存档的基因组、生物地球化学和环境时间序列数据，该项目寻求更新面向微生物的一维生物地球化学模型，以评估极地细菌群落的特定成员在模拟的生态过程中的不同贡献。应用机器学习建模技术，将上层海洋群落结构数据分割成不同的细菌生态型。预测的代谢信息和共同定位的生物地球化学速率测量将被用来确定群落类型之间关键的生理和功能差异。这些结果将通知对数值模拟框架内使用的细菌状态变量(S)的修改。然后将使用一系列模拟实验来比较机器学习综合模型框架和基础模型框架之间的模型技能，并探索可能改进北极气候变化预测的保真度的可能性。制作的模型的改编源代码将通过开源存档作为北极科学界的资源提供给人们。该项目由地球科学局共同资助，以支持地球科学领域的AI/ML进步。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The Arctic Ocean is undergoing rapid environmental change, with cascading effects on nearly all aspects of the polar marine ecosystem, including the abundance and community composition of marine microbes, microscopic organisms (i.e., bacteria and archaea) that play critical roles in the cycling of elements within the ice-ocean system. Numerical modeling is a critical method for testing ecological mechanisms and predicting the impacts of change. However, current approaches are typically not well resolved for bacterial diversity or activity rates, a source of uncertainty for future projections of critical marine ecosystem functions such as biological carbon drawdown. Through statistical exploration of sequence-based observations of the current Arctic microbial community, this research aims to transform our understanding of cellular environmental responses into a scale relevant for ecosystem processes and improve numerical modeling of microbial community structure and functional diversity. This project funds one post-doctoral scholar and supports undergraduate student training to build capacity in computational Arctic research. Leveraging publicly archived genomic, biogeochemical, and environmental time-series data from the central Arctic Ocean, this project seeks to update a microbial oriented, one-dimensional biogeochemical model to assess the variable contributions of specific members of the polar bacterial community in modeled ecological processes. Applying machine learning modeling techniques, upper ocean community structure data will be segmented into distinct bacterial ecotypes. Predicted metabolic information and co-located biogeochemical rate measurements will be used to identify critical physiological and functional differences among community types. These results will inform modifications to the bacterial state variable(s) used within the numerical modeling framework. A series of modeling experiments will then be used to compare model skill between the machine-learning integrated and base model frameworks and explore possible improvements to the fidelity of Arctic climate change predictions. Adapted source codes of the produced model will be made accessible through open-source archiving as a resource for the Arctic science community.This project is co-funded by the Directorate for Geosciences to support AI/ML advancement in the geosciences.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.