Modeling biophysical controls on hypoxia in a shallow estuary using a Bayesian mechanistic approach

This study describes development of a mechanistically parsimonious model to dynamically simulate bottom layer (subpycnocline) dissolved oxygen (BLDO) concentration in the Neuse River Estuary, USA (1997–2015). The approach embeds differential equations controlling May–October BLDO within a Bayesian framework, enabling rigorous uncertainty quantification considering prior knowledge and calibration to historical data. Model simulations explain 62% of variability in bimonthly mean BLDO observations. Results indicate that during July–August, 36% of BLDO is consumed meeting oxygen demand associated with seasonal primary production, while the rest is depleted meeting long-term oxygen demand (LTOD), associated with storage of organic matter in estuary sediments. Interannual LTOD variation is associated with November–April longitudinal velocities, suggesting elevated flushing in winter decreases oxygen demands in summer. Results also indicate that the system is more responsive to nutrient loading reductions than previously thought, though it may take multiple years to produce measurable declines in hypoxia due to hydro-meteorological variability. •We present a hybrid Bayesian-biophysical model of estuarine oxygen depletion.•The model characterizes drivers of oxygen dynamics across daily-yearly time scales.•Sediment and water-column respiration both contribute significantly to hypoxia.•Sediment respiration is inversely related to winter river flows in the Neuse Estuary.•Reducing riverine nutrients is effective in decreasing the number of hypoxic days.