A novel ensemble-based conceptual-data-driven approach for improved streamflow simulations

A novel ensemble-based conceptual-data-driven approach (CDDA) is developed where a data-driven model (DDM) used to “correct” the residuals from an ensemble of hydrological (HM) simulations. The CDDA respects processes via HM and it benefits DDM's ability simulate complex relationship between input variables. can accomodate any DDM, allowing for different configurations be tested. tested streamflow simulation in three Swiss catchments HM, HBV (Hydrologiska Byråns Vattenbalansavdelning), coupled with eight DDMs: Multiple Linear Regression, k Nearest Neighbours Second-Order Volterra Series Model, Artificial Neural Networks, two variants eXtreme Gradient Boosting (XGB) Random Forests (RF). proposed was able improve mean continuous ranked probability score by 16–29% over standalone HM. Since XGB RF demonstrated best performance, they are recommended simulating residuals.