Data Driven, Causal Discovery of the Effects of Climate Change, Ocean Acidification and Management in Fish and Invertebrate Stocks

Lead PI: Petros Koumoutsakos, Harvard University

Objectives

1. Conduct a comparative study of existing empirical models (EM), and ecosystem models such as the Ecopath with Ecosim (EwE) (for which we have significant experience) as well as ABMs for the spatiotemporal evolution of fish and invertebrate stocks.
2. Expand and develop novel ABMs with interaction parameters reverse engineered from coarse grained data.
3. Apply machine learning algorithms and causal inference techniques to ABMs and EMs to elucidate causal mechanisms.
4. Engage the local stakeholders in fisheries and fishermen communities in the Gulf of Maine

Methodology
Machine Learning, Causal inference, Computational Modeling, Hierrchical Bayesian Inference

Rationale
We will establish, through data driven computational models and AI, verified and causal relationships between environmental factors, fisheries management, and spatiotemporal distributions of fish and invertebrate stocks in the Gulf of Maine. We will investigate factors associated with climate change and ocean acidification, viewed as primary causes of variations in fish and invertebrate stocks (cod and lobster in particular), and examine how their management may be contributing to this situation. The rational understanding of the relative importance of each factor and their causal relationships will be a hallmark of scientific understanding and a potent enabler for policy makers and fisheries.