Dissertation Proposal Defense by Maria Cristina Perez
Advisor: Gavin Fay
Committee Members:
- Steve Cadrin (SMAST)
- Mazdak Tootkaboni (UMassD)
- Sarah Gaichas (Hydra LLC)
Abstract:
The scientific advice for fisheries management has traditionally relied on single-species models. This approach assumes that changes in abundance are primarily driven by fishing pressure, recruitment, and aggregate natural mortality, without explicitly accounting for interspecific interactions such as predation. Ignoring trophic and technical interactions can bias estimates of stock status and obscure trade-offs among conservation goals, fishery yield, and fleet performance. As a result, there is growing interest in ecosystem-based fisheries management approaches that explicitly recognize species interactions and fleet structure.
The dissertation combines simulation testing and machine learning emulation and scenario-based projections to improve the interpretation and usability of multispecies models for the Georges Bank region in the Northeast U.S. in a management context. focusing on a length-based multispecies and multifleet Model of Intermediate Complexity for Ecosystem Assessment. An Introductory Chapter reviews relevant literature and overviews the case study. Chapter 2 conducts simulation testing to assess how uncertainty in trophic interaction strength affects the ability to estimate state variables and key parameters such as fishing mortality, recruitment, biomass, and predation mortality. Chapter 3 develops machine-learning statistical emulators trained on outputs from a Management Strategy Evaluation conducted by the New England Fishery Management Council to rapidly reproduce biomass and catch dynamics under varying fishing conditions. Recent studies have shown that emulators can provide a computationally efficient way to explore sensitivity, uncertainty, and scenario space. Chapter 4 conducts long-term multispecies projections for a case system of ten Georges Bank fish species to explore how alternative patterns of fleet fishing mortality, fleet allocation, and relative preference for target species within fishing fleets shapes multispecies outcomes and trade-offs. The chapter also compares emulator-based projections to full multispecies projections for fleet-level fishing mortality scenarios to evaluate when emulators can reliably approximate ecosystem model behavior. Together, these chapters demonstrate how simulation testing, emulation, and targeted projections can be used to better understand uncertainty, expand scenario exploration, and improve the practical usability of multispecies fisheries models for management applications in complex, mixed fisheries.
SMAST West 204, 706 S. Rodney French Blvd, New Bedford 02744 and via Zoom
Callie Rumbut
c.rumbut@umassd.edu
https://umassd.zoom.us/j/91941653389?pwd=UY93xDKUDNCdZJaq2Hj0WWqb9SY51r.1