Applying community structure data and machine learning to improve microbial diversity in polar ecosystem models

Numerical modeling is a valuable tool for understanding ecosystem processes in the rapidly changing polar oceans. Due to their roles in biogeochemical cycling, climate change feedbacks, and the polar food web, it is increasingly important to explicitly quantify microbial contributions to modeled processes. However, current approaches are typically not informed by diversity data, despite clearly observed functional differences among microbial groups and increased ‘omics data availability. This results in a major discrepancy between observed and modeled biological complexity. Here, we highlight how machine learning, such as self-organizing maps and network analyses, can be used to segment the microbial community into functionally distinct ecotypes that can inform discrete variables in mechanistic single-column models. Specifically, we showcase a microbially-oriented regional test bed model that has been optimized for the western Antarctic Peninsula, but whose features can be altered for use in any region. For instance, we will use the biogeochemical, environmental, and genomic time-series data collected during the 2019-2020 MOSAiC Expedition to optimize this model for the warming central Arctic Ocean. Implementing our machine learning results, we will then test the variable contributions of specific archetypes within the microbial community to key ecological processes under climate change. We expect adding greater explicit functional diversity within this microbial-oriented numerical model will result in more accurate predictions of polar ecosystem function.

Presentation Preference: Either

Primary Presenter: Emelia Chamberlain, Woods Hole Oceanographic Institution (emelia.chamberlain@whoi.edu)

Authors:

Emelia Chamberlain, Woods Hole Oceanographic Institution  (emelia.chamberlain@whoi.eud)

Theodore Calianos, Woods Hole Oceanographic Institution  (theodore.calianos@whoi.edu)

Elizabeth Connors, Scripps Institution of Oceanography, University of California San Diego  (econnors@ucsd.edu)

Heather Kim, Woods Hole Oceanographic Institution  (hkim@whoi.edu)