321 / 2024-02-29 15:31:17

Exploring the effects of modifying stratification reality and biogeochemical model parameters on hypoxia using DN-4DVar

4D-Var,data assimilation,Hypoxia,Osaka Bay,Parameter optimization

The purpose of this study is to evaluate the influence and performance of stratification and biogeochemical model parameters for enhanced accuracy in the simulation of the spatial distribution of hypoxia in estuaries. Predicting the spatial distribution of hypoxia is critical for the aquatic environment of Osaka Bay, given its substantial inflow of freshwater and nutrients. Although density stratification dominates the scale of hypoxia in estuaries, accurately simulating such unsteady and spatially varying density structures using ocean models is challenging. In addition, because the parameters of biogeochemical models are subject to uncertainty, the model requires parameterization procedures. These problems have the potential to cause complementation errors, and errors in the physical field can be compensated by over-modifying the model parameters. This limitation may result in unrealistic predictions, particularly during time extrapolation including future forecasting. The model parameters and density distributions must be evaluated independently to ensure realism. We applied Regional Ocean Modeling Systems (ROMS) to simulate hydrodynamics and hypoxia and confirmed the performance of the model by comparing hourly monitoring data in the bay. The target period was one month of August 2012. The monitored water temperature, salinity, dissolved oxygen, and chlorophyll content were then assimilated using a dual-number-based four-dimensional variational data assimilation (DN-4DVar) method to correct the physical field and optimize the model parameters. A new parameter estimation term was added to the 4DVar evaluation function to achieve the simultaneous optimization of the physical field and model parameters. We focused on the role of these uncertainty factors in accurately simulating the hypoxia size. Four data assimilation experiments were performed by correcting the initial values of density, biogeochemical state variables, and/or model parameters. Correction of the initial values of density improved the horizontal extent of hypoxia, while optimization of the biogeochemical model parameters related to oxygen consumption by remineralization improved the thickness of hypoxia. The findings also highlight the uncertainty in parameters associated with primary production and underscore the modeling incompleteness related to the surface mixing layer during runoff. This study advances our understanding of hypoxic dynamics in the bay and holds broader implications for ecosystem management and climate change research.