LAUSR

Informing fishery management decisions using coupled socio-ecological systems (CSES) models requires model construction that captures the systems interactions with high precision. Ecological uncertainty in fishery models is easily reduced using existing scientific literature, but social drivers are often poorly defined or understood. The lack of knowledge about fishermen behavior results in inaccurate models of questionable utility for fishery managers. We designed and constructed a high fidelity agent based model (ABM) using the socio-ecological framework that reduces social system uncertainty by capturing complex behaviors using data-driven bounded rationality and feedback. The resulting generalized ABM of CSES dynamics was instantiated to Pacific salmon fisheries at Kenai river in Upper Cook Inlet, Alaska. The data-driven model construction fuses multiple data-sets for classification of social and ecological fishery regimes into stochastic distributions; the agent behaviors were generalized by evolving parametrized equations using data-driven machine learning; multiple non-trivial metrics on multiple scales verified model's accuracy and predictive capacity. The verified model of CSES dynamics at the Kenai river revealed recent instability in the dipnet fishery coupled dynamics, historic instability in the drift gillnet fishery coupled dynamics due to a compensatory and aggressive fishing strategy, and in the future the model will be used for scenario-based studies to understand the outcomes of alternative management strategies.