Bering Sea | Climate and Oceanography DOCTORATE POSTER PRESENTATION Validating the ROMS-NPZ Bering 10k model with empirical zooplankton data Presenter: Genoa Sullaway , gsullaway@alaska.edu, University of Alaska Fairbanks Curry Cunningham , cjcunningham@alaska.edu David Kimmel , david.kimmel@noaa.gov, Alaska Fisheries Science Center, NOAA Fisheries James Thorson , james.thorson@noaa.gov, NOAA Fisheries Alaska Fisheries Science Center Darren Pilcher , darren.pilcher@noaa.gov, CICOES, Pacific Marine Environmental Laboratory
Earth systems models have evolved from simulating physical processes to including biogeochemical and biological processes. The added complexity and interdisciplinary nature of these broader models represent a massive effort to advance ocean and atmospheric modeling. Physical and biological ocean system models are increasingly utilized in species distribution modeling, short term forecasting, long-term scenario planning and adaptive resource management. Cross disciplinary collaborations help inform models with empirical data and balance the parameterization processes across computational and biological complexity. Regional Ocean Modeling-Nutrient Phytoplankton Zooplankton (ROMS- NPZ) models seek to simulate and forecast biological dynamics in the ocean, including zooplankton, whose production dynamics are tightly linked to oceanographic conditions and are a direct link to higher trophic levels. The Bering Sea 10k ROMS-NPZ model has zooplankton biomass hindcasts beginning in 1970 for five zooplankton species groups, the most taxonomically resolved NPZ model to date. However, hindcasts have not been validated in depth with empirical data. We used a survey replication method to compare the ROMS-NPZ model hindcasts to empirical zooplankton data. We found differences in absolute biomass values, in addition to relative variation in estimated phenology, annual and spatial variation across species group. Next, we constructed a hybrid spatial distribution model (H-SDM), which uses the ROMS hindcast and environmental information as covariates to quantify zooplankton biomass in the Eastern Bering Sea. We used cross validation to compare spatial and temporal prediction skill between the H-SDM, the ROMS model and empirical model. The H-SDM model, with ROMS-NPZ hindcast as a covariate, had the greatest spatial prediction skill, while the ROMS-NPZ hindcast had the best skill in forecasting short-term relative zooplankton biomass trends. This research highlights areas where ROMS-NPZ can be implemented in short-term adaptive management and areas where it can be improved to better reflect empirical data and ecological processes.
Alaska Marine Science Symposium 2023 112
Made with FlippingBook flipbook maker