Estuarine Modeling: Does a higher grid resolution improve model performance?

James J Pauer1, Timothy J Feist2, Wilson Melendez3, Xiaomi Zhang2 and John C Lehrter4, (1)US Environmental Protection Agency, ORD/NHEERL/MED, Grosse Ile, MI, United States, (2)Trinity Engineering Associates, Inc., Grosse Ile, MI, United States, (3)Computer Sciences Corporation, Grosse Ile, MI, United States, (4)US Environmental Protection Agency, ORD/NHEERL/GED, Gulf Breeze, FL, United States
Abstract:
Ecological models are useful tools to explore cause effect relationships, test hypothesis and perform management scenarios. A mathematical model, the Gulf of Mexico Dissolved Oxygen Model (GoMDOM), has been developed and applied to the Louisiana continental shelf of the northern Gulf of Mexico to address long-standing hypoxia issues. A computational grid of approximately 6 km x 6 km cells and 20 sigma layers for a total of 64,320 cells was used. GoMDOM was driven by hydrodynamics from the Navy’s NCOM model, and comprehensive kinetic equations were used to describe nutrient, phytoplankton and oxygen dynamics. The model was able to represent the observed salinity (an indication of transport accuracy), biogeochemical processes and the concentrations and metrics including hypoxic area for 2003 and 2006. However, the scale of a model may influence the model’s capabilities. Models with low grid resolution and large cells improve computational speed but may not completely capture system dynamics, while models with high grid resolution and small cells may be take a very long time to run, and require extensive computer resources, but better simulate horizontal or vertical gradients. In order to explore the impact of a higher resolution on the model performance, the model was also executed on a 2 km x 2 km modeling grid (558,880 total cells) using the same external boundaries, forcing functions and model coefficients as used in the calibrated 6km version. Here we compare the two modeling grid sizes in terms of how well they fit observations, describing biogeochemical processes, and predicting hypoxia under different nutrient scenarios.