A Simple Model for Predicting Hypoxic Events in a Tidal Estuary

Abstract Low dissolved oxygen concentration threatens the ecosystem services provided by estuaries. Low oxygen events frequently occur in water bodies heavily modified by human activities such as the Elbe passing through the city of Hamburg. In preparation for early warning systems, we developed along this river a simple auto-regressive statistical model to identify hypoxic events—days with mean oxygen below 4 mg L -1 . The model uses three predictors observed in preceding days: (1) daily mean oxygen concentration, (2) water-, and (3) air-temperature. Its performance evaluated at three monitoring stations improves with an extended observation window. The most parsimonious model, which balances accuracy and complexity, achieves high predictive skill: precision exceeds 90% of correctly predicted events at stations downstream of the port of Hamburg and 80% at upstream stations. By adjusting the forecast horizon and observation window, the approach can provide early warning up to seven days in advance, with precision above 50% in the more affected downstream locations. We discuss potential improvements of the methodology and suggest its transferability to other water bodies.