Parameter optimization and Bayesian inference in an estuarine eutrophication model of intermediate complexity


The Neuse River Estuary, North Carolina, has been experiencing severe consequences of eutrophication in recent years including excessive algal blooms, low levels of dissolved oxygen, declining shellfish populations, large fishkills, and outbreaks of toxic microorganisms. As in many other marine systems, nitrogen has been identified as the pollutant of concern in the estuary because it is believed to stimulate the excessive algal growth that is at the root of other ecological problems. A model incorporating the mechanisms of algal growth and nutrient consumption in the Neuse River Estuary is formulated mathematically and implemented in the computer software AQUASIM. Key model parameters of water quality interest are calibrated to observations of system variables by minimizing the sum of the squares of the weighted difference between actual measurements and simulated results. The calibrated model reproduces the observed seasonal patterns of key system variables and thus demonstrates a predictive capability that is of use to policy makers when they are making decisions for sustainable environmental management. As future work we will implement Bayesian parameter estimation, which would improve the robustness of decision support by accounting for parameter uncertainty using probability distributions. Eventually, our model will be linked with a Bayesian version of the SPARROW watershed model as a Bayesian Network to be used for developing an adaptive implementation modeling and monitoring strategy (AIMMS) for the Neuse River basin.