Data-based mechanistic modelling and river flow forecasting

摘要: Abstract The paper briefly reviews the topic of rainfall-flow modelling and inductive, Data-Based Mechanistic (DBM) approach to stochastic, dynamic systems. It then uses DBM methods investigate nonlinear relationship between daily rainfall flow in Leaf River, Mississippi, USA. Initially, recursive State-Dependent Parameter (SDP) estimation is used identify, non-parametric (graphical) terms, location nature 'effective rainfall' nonlinearity. Parameterization this nonlinearity optimization a constrained version resulting model allow for its interpretation hydrologically meaningful Transfer Function (SDTF) form. Finally, as basis design realtime forecasting using an optimized SDP Kalman Filter (SDPKF) engine that includes heteroscedastic measurement noise.