The relationship between bottom water dissolved oxygen concentration, vertical stratification, and temperature was investigated for the Neuse River estuary, North Carolina, a shallow, intermittently-mixed estuary using approximately 10 years of weekly/biweekly, mid-channel data. A generalized additive model (GAM) was used to initially explore the major relationships among observed variables. The results of this statistical model guided the specification of a process-based model of oxygen dynamics that is consistent with theory pet simple enough to be parameterized using available field data. The nonlinear optimization procedure employed allows for the direct estimation of microbial oxygen consumption and physical reoxygenation rates, including the effects of temperature and vertical stratification. These estimated rates may better represent aggregate system behaviour than closed chamber measurements made in the laboratory and in situ. The resulting model describes 79% of the variation in dissolved oxygen concentration and is robust when compared across separate locations and time periods. Model predictions suggest that the spatial extent and duration of hypoxia in the bottom waters of the Neuse are controlled by the balance between the net oxygen depletion rate and the frequency of vertical mixing events. During cool months, oxygen consumption rates remain low enough to keep oxygen concentration well above levels of concern even under extended periods of stratification. A concentration below 4 mg 1(-1) is only expected under extended periods without vertical mixing when bottom water temperature exceeds 15 degreesC, while a concentration below 2 mg 1(-1) is only expected when water temperature exceeds 20 degreesC. To incorporate thr effects of parameter uncertainty, model error, and natural variability on model prediction, we used Monte Carlo simulation to generate distributions for the predicted number of days of hypoxia during the summer season. The expected number of days with a dissolved oxygen concentration less than 4 mg 1(-1) is 46.8 with a standard deviation of 4.7, while 23.8 days are expected to have an oxygen concentration below 2 mg 1(-1) with a standard deviation of 4.2 days. When joined with models relating nutrient loading and productivity to benthic and pelagic respiration rates, this model will be useful for probabilistically predicting the impact of nutrient management on the frequency of low oxygen events. (C) 2001 Academic Press.
