Dissolved Oxygen Dynamics of a Stream: Model Discrimination and Estimation of Parameter Variability Using an Extended Kalman Filter
A model identification procedure is presented based on an extended Kaiman filter (EKF) applied to the oxygen mass balance equation of a stream. The procedure is used with data from a small Danish stream to discriminate among eight mathematical models of the photosynthesis-light (P-I) response of macrophytes in the stream. Each model was tested for adequacy using objective criteria based on the expected behavior of the innovations sequence (the time series of differences between predicted and observed oxygen concentrations) derived from the EKF. Temporal variation of the model parameters was also examined using the EKF. The maximum photosynthetic rate (Pm) varied slowly over the year but was essentially stationary within any three to six day period. The low light efficiency of photosynthesis (Eo) varied from day to day within any short period, but three to six day means of Eo were essentially stationary over the year. The ratio (Ik = Pm/Eo) was highly correlated with both short and long term variations in daily mean light intensity.