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

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